International Journal of

Molecular Sciences

Article

Development and Technical Validation of anImmunoassay for the Detection ofAPP669–711 (Aβ−3–40) in Biological Samples

Hans W. Klafki 1,* , Petra Rieper 1, Anja Matzen 2, Silvia Zampar 1 , Oliver Wirths 1 ,Jonathan Vogelgsang 1 , Dirk Osterloh 3, Lara Rohdenburg 1, Timo J. Oberstein 4 ,Olaf Jahn 5 , Isaak Beyer 6, Ingolf Lachmann 3, Hans-Joachim Knölker 6 and Jens Wiltfang 1,7,8

1 Department of Psychiatry and Psychotherapy, University Medical Center (UMG), Georg-August-University,D37075 Göttingen, Germany; petra.rieper@med.uni-goettingen.de (P.R.);silvia.zampar@med.uni-goettingen.de (S.Z.); oliver.wirths@medizin.uni-goettingen.de (O.W.);jonathan.vogelgsang@med.uni-goettingen.de (J.V.); lara.rohdenburg@stud.uni-goettingen.de (L.R.);Jens.Wiltfang@med.uni-goettingen.de (J.W.)

2 IBL International GmbH, Tecan Group Company, D-22335 Hamburg, Germany; Anja.Matzen@tecan.com3 Roboscreen GmbH, D-04129 Leipzig, Germany; dirk.osterloh@roboscreen.com (D.O.);

ingolf.lachmann@roboscreen.com (I.L.)4 Department of Psychiatry and Psychotherapy, Friedrich-Alexander-University of Erlangen-Nuremberg,

D-91054 Erlangen, Germany; Timo.Oberstein@uk-erlangen.de5 Max-Planck-Institute of Experimental Medicine, Proteomics Group, D-37075 Göttingen, Germany;

jahn@em.mpg.de6 Faculty of Chemistry, Technische Universität Dresden, D-01069 Dresden, Germany;

isaak.beyer@web.de (I.B.); hans-joachim.knoelker@tu-dresden.de (H.-J.K.)7 German Center for Neurodegenerative Diseases (DZNE), D-37075 Göttingen, Germany8 Neurosciences and Signaling Group, Institute of Biomedicine (iBiMED), Department of Medical Sciences,

University of Aveiro, 3810-193 Aveiro, Portugal* Correspondence: hans.klafki@med.uni-goettingen.de

Received: 7 August 2020; Accepted: 2 September 2020; Published: 8 September 2020�����������������

Abstract: The ratio of amyloid precursor protein (APP)669–711 (Aβ−3–40)/Aβ1–42 in blood plasma wasreported to represent a novel Alzheimer’s disease biomarker. Here, we describe the characterization of twoantibodies against the N-terminus of Aβ−3–x and the development and “fit-for-purpose” technical validationof a sandwich immunoassay for the measurement of Aβ−3–40. Antibody selectivity was assessed by capillaryisoelectric focusing immunoassay, Western blot analysis, and immunohistochemistry. The analyticalvalidation addressed assay range, repeatability, specificity, between-run variability, impact of pre-analyticalsample handling procedures, assay interference, and analytical spike recoveries. Blood plasma was analyzedafter Aβ immunoprecipitation by a two-step immunoassay procedure. Both monoclonal antibodies detectedAβ−3–40 with no appreciable cross reactivity with Aβ1–40 or N-terminally truncated Aβ variants. However,the amyloid precursor protein was also recognized. The immunoassay showed high selectivity for Aβ−3–40

with a quantitative assay range of 22 pg/mL–7.5 ng/mL. Acceptable intermediate imprecision of thecomplete two-step immunoassay was reached after normalization. In a small clinical sample, the measuredAβ42/Aβ−3–40 and Aβ42/Aβ40 ratios were lower in patients with dementia of the Alzheimer’s type than inother dementias. In summary, the methodological groundwork for further optimization and future studiesaddressing the Aβ42/Aβ−3–40 ratio as a novel biomarker candidate for Alzheimer’s disease has been set.

Keywords: Alzheimer’s disease; biomarker; amyloid β; assay development; assay validation

Int. J. Mol. Sci. 2020, 21, 6564; doi:10.3390/ijms21186564 www.mdpi.com/journal/ijms

Int. J. Mol. Sci. 2020, 21, 6564 2 of 25

1. Introduction

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder that is responsible forthe majority of dementia cases [1]. Amyloid plaques containing aggregated forms of amyloid β

(Aβ) peptides represent one of the classical neuropathological hallmarks in AD brains [2,3]. Aβ isgenerated from the amyloid precursor protein (APP) by consecutive proteolytic cleavages executed byβ- and γ-secretases [4]. Oligomeric aggregated forms of Aβ can impair synaptic function and haveneurotoxic properties, and are thus believed to play a critical role in AD pathogenesis (for a recentreview, see, for example, [5]). In recent years, a variety of N- and C-terminally truncated or modifiedversions of the canonical Aβ1–40 and Aβ1–42 species have been detected in amyloid plaques [6–8],and it appears that the exact length and amino sequence of both N- and C-termini affect aggregationtendency and neurotoxicity of the different Aβ-species [9,10]. For example, Aβ peptides startingwith pyroglutamate at position 3 of the Aβ sequence (AβN3pE–X) were shown to be abundant insenile amyloid plaques [11]. This posttranslational modification seems to increase the oligomerization,fibrillization, and cytotoxicity of Aβ [12–15]. Interestingly, a recent solid-state NMR spectroscopystudy suggested that the secondary structure of Aβ oligomers was only marginally affected bypyroglutamate-modification of the Aβ N-terminus [16]. Currently, no disease modifying treatmentsfor AD targeting the underlying pathophysiological processes are available. Biochemical and imagingbiomarkers of the pathological changes in AD are important for an early and differential dementiadiagnosis, in order to identify subjects at high risk in preclinical stages and for participant selection in thecontext of clinical trials [17,18]. Recently, the ratios of the concentrations of specific variants of amyloidβ

(Aβ) in human blood plasma were discovered to predict brain amyloid pathology with high accuracy orto differentiate Alzheimer’s dementia patients and subjects with dementia due to other reasons. Theseplasma Aβ ratios include Aβ42/Aβ40 [19–21], Aβ1–40/Aβ1–42, and APP669–711/Aβ1–42 [22,23]. APP669–711

is a proteolytic fragment of the amyloid precursor protein (APP) that was identified in human bloodplasma by immunoprecipitation followed by mass spectrometry [24]. The APP669–711 peptide startsthree amino acids N-terminal to the canonical Aβ amino-terminus (Asp1) and encompasses the aminoacid sequence of an N-terminally elongated Aβ−3–40 peptide. Neuronal BE(2)-C cells and SH-SY5Ycells overexpressing APP were shown to be able to generate Aβ−3–40 in cell culture [23,25], and itappears that its cellular production occurs independently of BACE1 [25]. The self-assembly tendencyof Aβ−3–40 was found to be lower than that of Aβ1–42 [23]. To the best of our knowledge, N-terminallyelongated Aβ−3–x peptides have not been identified in appreciable amounts in amyloid plaques, thustheir potential significance in AD pathophysiology is currently unclear. The Aβ−3–40 (APP669–711)concentration in human blood plasma was reported to be essentially unchanged between subjects withor without amyloid-β positron-emission tomography (PET) imaging evidence of cerebral amyloid β

accumulation. However, it was found to be a good reference to accentuate the selective pathologicaldecrease in plasma Aβ1–42 in brain amyloid positive individuals. The plasma concentration ratioAPP669–711/Aβ1–42 (Aβ−3–40/Aβ1–42) clearly outperformed plasma Aβ1–42 as a biomarker of brainamyloid deposition [22,23]. We report here on the characterization of a novel monoclonal antibodythat was raised against the N-terminus of Aβ−3–x and on the development and technical validationof a novel immunoassay for the measurement of Aβ−3–40 in biological samples. Finally, the relativelevels of Aβ−3–40, Aβ40, and Aβ42 in human blood plasma in a small clinical sample were measuredafter pre-concentration by semi-automated Aβ magnetic bead immunoprecipitation. We observedstatistically significant differences between patients with dementia of the Alzheimer’s type (AD-D)(n = 23) and dementia due to other reasons (OD) (n = 17) regarding Aβ42 and the concentration ratiosAβ42/Aβ40 and Aβ42/Aβ−3–40.

2. Results

2.1. Antibody Characterization

An overview of APP- and Aβ-related peptides employed in this study and the nomenclature usedare shown in Figure 1.

Int. J. Mol. Sci. 2020, 21, 6564 3 of 25Int. J. Mol. Sci. 2020, 21, x FOR PEER REVIEW 3 of 26

Figure 1. Amino acid sequences and numbering of the Aβ variants and amyloid precursor protein

(APP)-related peptides that were studied. On top, a schematic diagram of the primary structure of the

human APP770 isoform is shown. The Aβ peptide region, sAPPβ, and the APP intracellular domain

(AICD) are indicated by different colours. Partial amino acid sequences of wild type APP770 and a

mutant APP are shown in the middle panel. The positions of the Swedish (K670N/M671L) double

mutation as well as the I716V (London) and V717I (Florida) mutations are highlighted in red colour. In

the lower panel, the amino acid sequences and nomenclature of the 11 different synthetic peptides

employed in this study are summmarized.

The selectivity of the purified monoclonal antibody (mAb) 101-1-1 was assessed by capillary

isoelectric focusing (CIEF) immunoassay, as described previously [8,26]. MAb 101-1-1 showed a strong

signal with synthetic Aβ−3–40 (100 ng/mL), but not with Aβ1–40, Aβ2–40, Aβ3–40, AβN3pE–40, Aβ4–40, or Aβ5–40.

MAb 101-1-1 also recognized the antigen VKMDAEFRC coupled to bovine serum albumin (BSA) (Figure

2).

Figure 1. Amino acid sequences and numbering of the Aβ variants and amyloid precursor protein(APP)-related peptides that were studied. On top, a schematic diagram of the primary structure of thehuman APP770 isoform is shown. The Aβ peptide region, sAPPβ, and the APP intracellular domain(AICD) are indicated by different colours. Partial amino acid sequences of wild type APP770 and amutant APP are shown in the middle panel. The positions of the Swedish (K670N/M671L) doublemutation as well as the I716V (London) and V717I (Florida) mutations are highlighted in red colour.In the lower panel, the amino acid sequences and nomenclature of the 11 different synthetic peptidesemployed in this study are summmarized.

The selectivity of the purified monoclonal antibody (mAb) 101-1-1 was assessed by capillaryisoelectric focusing (CIEF) immunoassay, as described previously [8,26]. MAb 101-1-1 showed a strongsignal with synthetic Aβ−3–40 (100 ng/mL), but not with Aβ1–40, Aβ2–40, Aβ3–40, AβN3pE–40, Aβ4–40, orAβ5–40. MAb 101-1-1 also recognized the antigen VKMDAEFRC coupled to bovine serum albumin(BSA) (Figure 2).

Essentially identical electropherograms indicating excellent selectivity for Aβ−3–40 were obtainedwith mAb 14-2-4 (anti Aβ−3–X, IBL International, Hamburg, Germany) (Figure S1). In the next step, theselectivity of mAb101-1-1 in Western blot detection of Aβ peptide variants separated under stronglydenaturing conditions by urea-SDS-polyacrylamide gel electrophoresis (urea SDS-PAGE) was assessed.Again, mAb 101-1-1 showed excellent selectivity for synthetic Aβ−3–40. Under the tested conditions,no signals were observed with Aβ variants starting with Asp (1), Ala (2), Glu (3), pyroGlu (3), Phe(4), or Arg (5) (Figure 3a,b). Very similar results were obtained with mAb 14-2-4 (Figure S2). Toevaluate whether the monoclonals 101-1-1 and 14-2-4 require the free N-terminal valine residue oralso recognize APP or APP fragments starting N-terminal to the canonical Aβ sequence, we testedrecombinant sAPPα and cell culture supernatants of H4 cells transfected with human wild type APP751or Swedish mutant APP751SWE (K670N/M671L). Aliquots of cell culture supernatants and recombinantsAPPα were separated on a 4–12% Bis-Tris gradient gel, blotted onto PVDF (polyvinylidene difluoridemembrane), and probed with mAb 101-1-1 or mAb 14-2-4 in combination with peroxidase labeled goatanti mouse IgG secondary antibody. For comparison, additional parallel blots were probed with mAb22C11 (anti-APP) and mAb 1E8 (anti Aβ N-terminus). Both mAb 101-1-1 and 14-2-4 clearly detected

Int. J. Mol. Sci. 2020, 21, 6564 4 of 25

recombinant sAPPα and secrected sAPP in cell culture supernatant of H4 cells overexpressing humanwild type APP751, but not in conditioned medium of H4 cells overexpressing Swedish mutant APP(Figure 3c–f).Int. J. Mol. Sci. 2020, 21, x FOR PEER REVIEW 4 of 26

Figure 2. Monoclonal antibody 101-1-1 recognizes Aβ−3–40 with high selectivity on capillary isoelectric

focusing immunoassay. A series of synthetic Aβ peptides with different N-termini was separated by

isoelectric focusing in microcapillaries, immobilized by a photochemical reaction, and probed with

(a) mAb 6E10 (2 µg/mL) or (b) mAb 101-1-1 (2.1 µg/mL). Chemiluminescent detection was achieved

with biotinylated goat-anti mouse IgG antibody in combination with streptavidin-HRP (streptavidin

horseradish peroxase conjugate) and chemiluminescent substrate. The specific Aβ variants loaded in

the different capillaries as single peptides or in mixtures are indicated. Aβ1–40, Aβ2–40, Aβ3–40, Aβ4–40,

Aβ5–40, and Aβ−3–40 were loaded at a concentration of 100 ng/mL. The tested concentration of AβN3pE–40

and VKMDAEFRC-bovine serum albumin (BSA) was 200 ng/mL.

Essentially identical electropherograms indicating excellent selectivity for Aβ−3–40 were obtained

with mAb 14-2-4 (anti Aβ−3–X, IBL International, Hamburg, Germany) (Figure S1). In the next step, the

selectivity of mAb101-1-1 in Western blot detection of Aβ peptide variants separated under strongly

denaturing conditions by urea-SDS-polyacrylamide gel electrophoresis (urea SDS-PAGE) was assessed.

Again, mAb 101-1-1 showed excellent selectivity for synthetic Aβ−3–40. Under the tested conditions, no

signals were observed with Aβ variants starting with Asp (1), Ala (2), Glu (3), pyroGlu (3), Phe (4), or

Figure 2. Monoclonal antibody 101-1-1 recognizes Aβ−3–40 with high selectivity on capillary isoelectricfocusing immunoassay. A series of synthetic Aβ peptides with different N-termini was separated byisoelectric focusing in microcapillaries, immobilized by a photochemical reaction, and probed with(a) mAb 6E10 (2 µg/mL) or (b) mAb 101-1-1 (2.1 µg/mL). Chemiluminescent detection was achievedwith biotinylated goat-anti mouse IgG antibody in combination with streptavidin-HRP (streptavidinhorseradish peroxase conjugate) and chemiluminescent substrate. The specific Aβ variants loadedin the different capillaries as single peptides or in mixtures are indicated. Aβ1–40, Aβ2–40, Aβ3–40,Aβ4–40, Aβ5–40, and Aβ−3–40 were loaded at a concentration of 100 ng/mL. The tested concentration ofAβN3pE–40 and VKMDAEFRC-bovine serum albumin (BSA) was 200 ng/mL.

Int. J. Mol. Sci. 2020, 21, 6564 5 of 25

Int. J. Mol. Sci. 2020, 21, x FOR PEER REVIEW 5 of 26

Arg (5) (Figure 3a). Very similar results were obtained with mAb 14-2-4 (Figure S2). To evaluate whether

the monoclonals 101-1-1 and 14-2-4 require the free N-terminal valine residue or also recognize APP or

APP fragments starting N-terminal to the canonical Aβ sequence, we tested recombinant sAPPα and cell

culture supernatants of H4 cells transfected with human wild type APP751 or Swedish mutant

APP751SWE (K670N/M671L). Aliquots of cell culture supernatants and recombinant sAPPα were

separated on a 4–12% Bis-Tris gradient gel, blotted onto PVDF (polyvinylidene difluoride membrane),

and probed with mAb 101-1-1 or mAb 14-2-4 in combination with peroxidase labeled goat anti mouse

IgG secondary antibody. For comparison, additional parallel blots were probed with mAb 22C11

(anti-APP) and mAb 1E8 (anti Aβ N-terminus). Both mAb 101-1-1 and 14-2-4 clearly detected

recombinant sAPPα and secrected sAPP in cell culture supernatant of H4 cells overexpressing human

wild type APP751, but not in conditioned medium of H4 cells overexpressing Swedish mutant APP

(Figure 3c–f).

Figure 3. Western blot analysis of antibody selectivity. Upper panel: (a) The indicated synthetic Aβ

peptides with different N-termini were separated on a 12%T/5% C urea-SDS-polycacrylamide gel, blotted

onto PVDF, and probed with mAb 101-1-1 (1 µg/mL) in combination with goat anti-mouse

IgG-peroxidase conjugate (1:30,000 dilution in PBS-T (phosphate buffered saline containing 0.075%

Tween 20)). The peptide amounts loaded on the gel were 25 ng (Aβ4–40 and Aβ5–40) and 10 ng (remaining

Aβ peptides), respectively. (M) Protein ladder; (Aβ mix-1) mixture containing 10 ng of each of Aβ1–37, Aβ1–

38, Aβ1–39, Aβ1–40, and Aβ1–42; (Aβ mix-2) mixture containing 10 ng of each of Aβ1–37, Aβ1–38, Aβ1–39, Aβ1–40,

Aβ1–42, and Aβ−3–40. The image was recorded after 5 min exposure. Image display: high: 65,535; low: 0;

gamma: 0.6. (b) Reprobing of the same PVDF membrane without stripping with mAb 6E10 (1 µg/mL).

Image recorded after 3 min exposure. Image display: high: 65,535; low: 0; gamma 0.61. Lower panel:

Conditioned cell culture media of transfected H4 cells overexpressing Swedish mutant or wild type

human APP751 and recombinant sAPPα (stock solution prepared in Diluent-35) were separated by

SDS-polyacrylamide gel electrophoresis; blotted onto PVDF; and immuno-stained with (c) mAb 101-1-1,

(d) mAb 14-2-4, (e) mAb 22C11, and (f) mAb 1E8. (M) Protein ladder; (1) 0.5 µL of cell culture supernatant

H4APP751SWE; (2) 0.5 µL of cell culture supernatant H4APP751wt; and (3) 6.7 ng of recombinant sAPPα.

Blot exposure times: mAb 101-1-1: 15 min, mAb 14-2-4: 6 min, mAb 22C11: 2 min, and mAb 1E8: 6 min.

Image display (all four blots): max: 65,535; min: 0, gamma: 0.6). * Aβ (tentative assignment).

Figure 3. Western blot analysis of antibody selectivity. Upper panel: (a) The indicated synthetic Aβ

peptides with different N-termini were separated on a 12%T/5% C urea-SDS-polycacrylamide gel,blotted onto PVDF, and probed with mAb 101-1-1 (1 µg/mL) in combination with goat anti-mouseIgG-peroxidase conjugate (1:30,000 dilution in PBS-T (phosphate buffered saline containing 0.075%Tween 20)). The peptide amounts loaded on the gel were 25 ng (Aβ4–40 and Aβ5–40) and 10 ng (remainingAβ peptides), respectively. (M) Protein ladder; (Aβ mix-1) mixture containing 10 ng of each of Aβ1–37,Aβ1–38, Aβ1–39, Aβ1–40, and Aβ1–42; (Aβ mix-2) mixture containing 10 ng of each of Aβ1–37, Aβ1–38,Aβ1–39, Aβ1–40, Aβ1–42, and Aβ−3–40. The image was recorded after 5 min exposure. Image display:high: 65,535; low: 0; gamma: 0.6. (b) Reprobing of the same PVDF membrane without stripping withmAb 6E10 (1 µg/mL). Image recorded after 3 min exposure. Image display: high: 65,535; low: 0; gamma0.61. Lower panel: Conditioned cell culture media of transfected H4 cells overexpressing Swedish mutantor wild type human APP751 and recombinant sAPPα (stock solution prepared in Diluent-35) wereseparated by SDS-polyacrylamide gel electrophoresis; blotted onto PVDF; and immuno-stained with (c)mAb 101-1-1, (d) mAb 14-2-4, (e) mAb 22C11, and (f) mAb 1E8. (M) Protein ladder; (1) 0.5 µL of cellculture supernatant H4APP751SWE; (2) 0.5 µL of cell culture supernatant H4APP751wt; and (3) 6.7 ngof recombinant sAPPα. Blot exposure times: mAb 101-1-1: 15 min, mAb 14-2-4: 6 min, mAb 22C11:2 min, and mAb 1E8: 6 min. Image display (all four blots): max: 65,535; min: 0, gamma: 0.6). * Aβ

(tentative assignment).

The recognition of wild type APP, but not APP carrying the Swedish double mutation KM/NL(670/671) by mAb 101-1-1 was further confirmed by immunohistochemistry (Figure 4). MAb 101-1-1clearly stained cellular APP in brain slices from APPLd transgenic mice, but not in brain slices from5XFAD transgenic mice. The latter express human APP and Presenilin 1 genes with a total of fivemutations linked to AD, including the APP SWE double mutation. Double staining of brain slices ofAPPLd mice with anti Aβ42 (mAb D3E10) and mAb 101-1-1 indicated Aβ42 to be located in amyloidplaque cores, as expected. These were not recognized by mAb 101-1-1. Instead, mAb 101-1-1 signalswere observed in close vicinity to the extracellular plaque cores, presumably representing APPaccumulation within dystrophic neurites.

Int. J. Mol. Sci. 2020, 21, 6564 6 of 25

Int. J. Mol. Sci. 2020, 21, x FOR PEER REVIEW 6 of 26

The recognition of wild type APP, but not APP carrying the Swedish double mutation KM/NL

(670/671) by mAb 101-1-1 was further confirmed by immunohistochemistry (Figure 4). MAb 101-1-1

clearly stained cellular APP in brain slices from APPLd transgenic mice, but not in brain slices from

5XFAD transgenic mice. The latter express human APP and Presenilin 1 genes with a total of five

mutations linked to AD, including the APP SWE double mutation. Double staining of brain slices of

APPLd mice with anti Aβ42 (mAb D3E10) and mAb 101-1-1 indicated Aβ42 to be located in amyloid

plaque cores, as expected. These were not recognized by mAb 101-1-1. Instead, mAb 101-1-1 signals

were observed in close vicinity to the extracellular plaque cores, presumably representing APP

accumulation within dystrophic neurites.

Figure 4. Immunohistochemical staining of Alzheimer’s disease (AD) transgenic mouse models. (a)

Staining using an antibody directed against the carboxyterminus of APP (APP C-term) showed

abundant cellular and neuritic immunoreactivity in 5XFAD and APPLd mice (left). In contrast, 101-1-1

showed an abundant staining only in APPLd mice harboring APP with an aminoterminal wildtype

(K670–M671) sequence, but a complete lack of immunoreactivity in 5XFAD mice overexpressing

APP with the Swedish mutation (N670–L671) (right). (b) Double-staining using an Aβ42-specific

antibody (D3E10, green) and 101-1-1 (red) revealed abundant Aβ immunoreactivity in the central

amyloid plaque cores, while the 101-1-1 signal was restricted to surrounding dystrophic neurites.

Scale bar: 100 µm.

Next, the suitability of mAb 101-1-1 for magnetic bead immunoprecipitation (IP) after covalent

coupling to Dynabeads M270 Epoxy was tested. Western blot analysis revealed successful IP of sAPP

from conditioned media of H4APP751wt cells (Figure S3). After extended exposure, a faint peptide band

co-migrating with synthetic Aβ peptides Aβ−3–40 and Aβ1–40 was also detected.

2.2. Immunoassay Development and Technical Validation

A highly selective sandwich immunoassay for measuring Aβ−3–40 (APP669–711) was developed on

the Mesoscale Discovery (MSD) technology platform employing 96-well small spot streptavidin

pre-coated plates. Aliquots of mAb 101-1-1 and the anti Aβ40 C-terminal mAb 280F2 were conjugated

Figure 4. Immunohistochemical staining of Alzheimer’s disease (AD) transgenic mouse models.(a) Staining using an antibody directed against the carboxyterminus of APP (APP C-term) showedabundant cellular and neuritic immunoreactivity in 5XFAD and APPLd mice (left). In contrast, 101-1-1showed an abundant staining only in APPLd mice harboring APP with an aminoterminal wildtype(K670–M671) sequence, but a complete lack of immunoreactivity in 5XFAD mice overexpressing APPwith the Swedish mutation (N670–L671) (right). (b) Double-staining using an Aβ42-specific antibody(D3E10, green) and 101-1-1 (red) revealed abundant Aβ immunoreactivity in the central amyloid plaquecores, while the 101-1-1 signal was restricted to surrounding dystrophic neurites. Scale bar: 100 µm.

Next, the suitability of mAb 101-1-1 for magnetic bead immunoprecipitation (IP) after covalentcoupling to Dynabeads M270 Epoxy was tested. Western blot analysis revealed successful IP of sAPPfrom conditioned media of H4APP751wt cells (Figure S3). After extended exposure, a faint peptideband co-migrating with synthetic Aβ peptides Aβ−3–40 and Aβ1–40 was also detected.

2.2. Immunoassay Development and Technical Validation

A highly selective sandwich immunoassay for measuring Aβ−3–40 (APP669–711) was developedon the Mesoscale Discovery (MSD) technology platform employing 96-well small spot streptavidinpre-coated plates. Aliquots of mAb 101-1-1 and the anti Aβ40 C-terminal mAb 280F2 were conjugated(initially on a small scale) with biotin or SULFO-TAG to serve as capture or detection antibodies,respectively. After blocking unspecific binding sites, selected wells of a 96-well small spot streptavidinplate were coated with biotinylated mAb 101-1-1 capture antibody. A series of fourfold dilutionsof synthetic Aβ−3–40 in Diluent-35 (1500 pg/mL, 375 pg/mL, 93.75 pg/mL, 23.44 pg/mL, 5.86 pg/mL,1.46 pg/mL, 0.37 pg/mL, and 0 pg/mL) was assessed. With both SULFO-TAG Aβ40 detection antibody(MSD) and SULFO-TAG mAb 280F2, reasonable dose–response curves were obtained. The lower limitsof detection (LLODs) were calculated by the MSD Discovery Workbench software as the smallestanalyte concentration generating a signal three standard deviations (3 × SD) above the zero calibrator(blank) (minimum error estimates activated). The observed LLODs of 24.6 pg/mL (SULFO-TAG Aβ40)and 22.7 pg/mL (SULFO-TAG 280F2) were very similar. The flipped assay design with biotinylatedmAb 280F2 serving for capture and SULFO-TAG 101-1-1 for detection produced substantially lower

Int. J. Mol. Sci. 2020, 21, 6564 7 of 25

signals and a considerably higher calculated LLOD of 72.1 pg/mL (Figure 5a). Thus, we decided toproceed with assay development using biotinylated mAb 101-1-1 for capture and SULFO-TAG Aβ40

for detection. Next, a fresh lot of biotinylated mAb 101-1-1 was prepared on a larger scale, starting from1 mg of purified mAb 101-1-1. The mole to mole ratio of incorporated biotin to protein was estimated byHABA (4-hydroxyazobenzene-2-carboxylic acid) assay for biotin quantitation and BCA (bicinchoninicacid) assay for protein quantitation. We calculated incorporation of approximately 2.9 moles of biotinper mole of protein and IgG recovery of approximately 90%. To optimize the detection of Aβ−3–40,different concentrations of biotinylated mAb 101-1-1 capture antibody during plate coating were tested(capture antibody titration). We observed an appreciable improvement in the detection sensitivity byincreasing the concentration of the capture antibody from 0.5 µg/mL to 1 µg/mL and 2 µg/mL. Furtherincreasing the capture antibody concentration to 4 µg/mL had a negligible additional effect (Figure 5b).At this stage and based on the results obtained so far, we drafted a first version of a standard operatingprocedure (SOP, see below, Materials and Methods section) to be evaluated in detail in a partial “fit forpurpose” assay validation campaign. A cartoon displaying the assay principle is shown in Figure S4.

Int. J. Mol. Sci. 2020, 21, x FOR PEER REVIEW 8 of 26

prepared in H2O, and further dilutions were done in Diluent-35. Fourfold serial dilutions of the

D7-Calibrator starting with 1029 pg/mL produced a reasonable standard curve with an LLOD of 11.2

pg/mL (Figure 5d). To compare dose–response curves, taking into account the difference in the

molecular masses of Aβ−3–40 (Mr = 4685.4) and the D7-Calibrator (Mr = 2571), the x-axis of the plot

shown in Figure 5d was scaled in moles/L. At equal molar concentrations, Aβ−3–40 produced slightly

higher signals than the D7-Calibrator. Nevertheless, the findings suggest that the D7-Calibrator

provides an optional alternative standard/calibrator to the full length Aβ−3–40 peptide.

Figure 5. Assessment of different antibody combinations, capture antibody titration, quantitative assay

range, and comparison of alternative calibrators. (a) Comparison of sandwich immunoassay variations

employing mAb 101-1-1-biotin or 280F2-biotin for capture and SULFO-TAG Aβ40, SULFO-TAG 280F2, or

SULFO-TAG 101-1-1 for detection. The tested antibody combinations are indicated. (b) lower limit of

detection (LLOD) and lower limit of quantification (LLOQ) of the Aβ−3–40 immunoassay as functions of

the 101-1-1-biotin capture antibody concentration. (c) Determination of the quantitative assay range.

Upper limit of quantification (ULOQ) and LLOQ were calculated from the signals obtained with an

extended fourfold calibrator dilution series plus blank (zero calibrator). (d) Comparison of dose–response

curves obtained with synthetic Aβ−3–40 and D7-Calibrator.

2.2.3. Technical Intra-Assay Repeatability and Spatial Bias

To check for potential positional effects (spatial bias), synthetic Aβ−3–40 was diluted in Diluent-35

to a final concentration of 469 pg/mL and measured in 24 technical replicates distributed across one

96-well assay plate. The coefficient of variation in terms of the measured signals was 7.6%. The mean

calculated concentration was 485.8 pg/L and the coefficient of variation (%CV) between calculated

Aβ−3–40 concentrations in the 24 replicates was 4.7%. Thus, in general, the technical repeatability of

the assay was good. However, in position H12 (lower right corner of the assay plate), the measured

signal was 23% lower than the mean signal. To exclude a systematic positional effect (spatial bias) of

our system, on a second assay plate, we tested 11 technical replicates of a corresponding Aβ−3–40

dilution. The observed %CVs were 11.1% (measured signal) and 7.1% (calculated concentrations).

Figure 5. Assessment of different antibody combinations, capture antibody titration, quantitative assayrange, and comparison of alternative calibrators. (a) Comparison of sandwich immunoassay variationsemploying mAb 101-1-1-biotin or 280F2-biotin for capture and SULFO-TAG Aβ40, SULFO-TAG 280F2, orSULFO-TAG 101-1-1 for detection. The tested antibody combinations are indicated. (b) lower limit ofdetection (LLOD) and lower limit of quantification (LLOQ) of the Aβ−3–40 immunoassay as functionsof the 101-1-1-biotin capture antibody concentration. (c) Determination of the quantitative assay range.Upper limit of quantification (ULOQ) and LLOQ were calculated from the signals obtained with anextended fourfold calibrator dilution series plus blank (zero calibrator). (d) Comparison of dose–responsecurves obtained with synthetic Aβ−3–40 and D7-Calibrator.

2.2.1. Lower Limit of Detection, Lower Limit of Quantification, and Upper Limit of Quantification

The data analysis in each assay run included standard curve calculations with the MSD DiscoveryWorkbench 4.0.12 software by four-parameter logistic curve fitting of raw signals obtained with seven

Int. J. Mol. Sci. 2020, 21, 6564 8 of 25

or more fourfold serial dilutions of the synthetic Aβ−3–40 calibrator peptide plus a zero calibrator(blank). The lower limit of detection (LLOD) was calculated for each single assay plate as the lowestanalyte concentration generating a signal three SDs above the lowest standard/blank with the option“use minimum error estimates” activated. The lower limit of quantification (LLOQ) was defined as thelowest analyte concentration producing a signal 10 SDs above the lowest standard/blank. Mean LLODand LLOQ calculated from n = 5 assay runs within the validation study were 11.04 pg/mL ± 0.44 and22.24 pg/mL ± 1.36 (mean ± SD), respectively. To assess the upper limit of quantification (ULOQ),on a single assay plate, an extended fourfold calibrator dilution series including ten fourfold serialdilutions starting at 120 ng/mL plus a zero calibrator/blank was analyzed. The ULOQ was definedas the highest calibrator concentration that showed < 20% coefficient of variation (CV) between twotechnical replicates (regarding both signal and calculated concentration) and 85–115% recovery (i.e., thecalculated concentration had to be in the range of 85–115% of the theoretical calibrator concentration).The observed ULOQ in this experiment was 7.5 ng/mL (Figure 5c).

2.2.2. Alternative Calibrator

Aβ peptides are notoriously prone to aggregation. We thus evaluated a synthetic molecule(D7-Calibrator) comprising the Aβ−3–X N-terminal amino acid sequence VKMDAEFRH followed by 5xPEG2 (8-Amino-3,6-dioxaoctanoic acid) and the Aβ40 C-terminus GLMVGGVV (Mr: 2571) (Biosyntan,Berlin, Germany) as a potential alternative calibrator. The D7-Calibrator was designed to provideimproved water-solubility and to be less prone to aggregation or adsorption. The physicochemicalproperties have not been studied in detail. A stock solution (3.05 mg/mL) was prepared in H2O, andfurther dilutions were done in Diluent-35. Fourfold serial dilutions of the D7-Calibrator startingwith 1029 pg/mL produced a reasonable standard curve with an LLOD of 11.2 pg/mL (Figure 5d).To compare dose–response curves, taking into account the difference in the molecular masses ofAβ−3–40 (Mr = 4685.4) and the D7-Calibrator (Mr = 2571), the x-axis of the plot shown in Figure 5dwas scaled in moles/L. At equal molar concentrations, Aβ−3–40 produced slightly higher signals thanthe D7-Calibrator. Nevertheless, the findings suggest that the D7-Calibrator provides an optionalalternative standard/calibrator to the full length Aβ−3–40 peptide.

2.2.3. Technical Intra-Assay Repeatability and Spatial Bias

To check for potential positional effects (spatial bias), synthetic Aβ−3–40 was diluted in Diluent-35to a final concentration of 469 pg/mL and measured in 24 technical replicates distributed across one96-well assay plate. The coefficient of variation in terms of the measured signals was 7.6%. The meancalculated concentration was 485.8 pg/L and the coefficient of variation (%CV) between calculatedAβ−3–40 concentrations in the 24 replicates was 4.7%. Thus, in general, the technical repeatability of theassay was good. However, in position H12 (lower right corner of the assay plate), the measured signalwas 23% lower than the mean signal. To exclude a systematic positional effect (spatial bias) of oursystem, on a second assay plate, we tested 11 technical replicates of a corresponding Aβ−3–40 dilution.The observed %CVs were 11.1% (measured signal) and 7.1% (calculated concentrations). Importantly,on this assay plate, the measured signal in position H12 was 10.3% above the mean signal, suggestingrandom error and arguing against systematic spatial bias. Heat maps indicating the distribution of themeasured signals are shown in Figure S5.

2.2.4. Cross Reactivity/Assay Specificity

Next, we assessed the cross reactivity of the sandwich immune assay with related Aβ peptidevariants and sAPPα. To that end, Aβ1–40, Aβ1–42, Aβ−3–38, and Aβ−3–42 were tested at concentrationsof 4 ng/mL and 20 ng/mL and recombinant sAPPα at a concentration of 320 ng/mL. Aβ1–40, Aβ1–42, andAβ−3–38 measurements at both tested concentrations were below the detection range of the assay plate.Aβ−3–42 produced signals within the detection range. The calculated concentrations were 12.5 pg/mLwhen 4 ng/mL of Aβ−3–42 was applied and 51.6 pg/mL when 20 ng/mL of Aβ−3–42 was analyzed.

Int. J. Mol. Sci. 2020, 21, 6564 9 of 25

Thus, approximately 0.3% of the Aβ−3–42 peptide was recognized in the Aβ−3–40 assay, indicating across reactivity <0.5%. The signal observed with 320 ng/mL of recombinant sAPPα was below thedetection range of the assay (Table 1).

Table 1. Cross reactivity of the Aβ−3–40 immunoassay.

Sample Concentration (pg/mL) Signal Calculated Concentration (pg/mL) % Recognized

Aβ1–40 4000 93 below detection range n.a.Aβ1–40 20,000 79 below detection range n.a.

Aβ−3–38 4000 93 below detection range n.a.Aβ−3–38 20,000 92 below detection range n.a.Aβ−3–42 4000 139 12.5 0.31Aβ−3–42 20,000 594 51.6 0.26Aβ1–42 4000 110 below detection range n.a.Aβ1–42 20,000 85 below detection range n.a.sAPPα 320,000 91 below detection range n.a.

zero calibrator 0 88 n.a.

n.a.: not applicable, APP: amyloid precursor protein.

2.2.5. Between-Run Variability/Inter-Assay Variance of the Aβ−3–40 Immunoassay

The initial core technical assay validation campaign included five independent runs of theAβ−3–40 chemiluminescence immunoassay that were executed within eight weeks. On each of the fiveassay plates, an aliquot of a quality control (QC) sample was measured in four technical replicates.The primary intended application of the novel assay was to measure Aβ−3–40 after pre-analyticalenrichment of Aβ peptides from EDTA-blood plasma by magnetic bead immunoprecipitation (IP) [21].A QC-IP-eluate was prepared by running several IPs in parallel from pooled human blood plasmaand combining the eluates after 4.8-fold dilution with Diluent-35 to a single QC-IP-eluate. Aliquotsfor single use were stored at −80 ◦C until analysis. The results obtained with the QC-IP-eluate in fiveindependent assay runs are summarized in Table 2. The electro-chemiluminescence signals observedin the five different experiments (means of four technical replicates on each assay plate) ranged from6755 to 8216, with a coefficient of variation of 8.2%. The mean calculated concentration of Aβ−3–40 inthe QC-IP-eluate was 217 pg/mL, with a coefficient of variation of 16.3%, which was outside of thepredefined acceptance range of 15% CV (Table 2). The relatively large scatter in terms of the calculatedconcentration of Aβ−3–40 was apparently caused by assay run #4, which seemed to differ substantiallyfrom the remaining four experiments and was identified as an outlier by Grubbs test (α = 0.05). Notably,the serial calibrator dilutions for assay run #4 had been prepared from a higher starting concentration,and thus following a slightly different experimental protocol than for the remaining experiments.This was done in order to allow for determination of the ULOQ on assay plate #4 (see above). Excludingplate #4 from the analysis reduced the CV of the measured concentrations to 4.7%. Taken together,the findings suggest that a strict standard operation procedure (SOP) covering all experimental stepsincluding the preparation of calibrator dilutions needs to be implemented. Furthermore, a suitable QCsample should be included to allow for normalization between assay plates (see below).

Table 2. Day-to-day variance of the Aβ−3–40 assay.

Assay Run Mean Signal Concentration (pg/mL) *

1 6755 193.602 7400 195.493 7022 204.524 7924 279.165 8216 214.36

Mean 7463.40 217.43SD 608.12 35.48

%CV 8.15 16.32

* on each assay plate, an aliquot of the same original diluted quality control (QC) immunoprecipitation (IP) eluatewas measured in four technical replicates. The indicated concentrations refer to the mean calculated concentrationin the diluted IP eluate without correction for dilution.

Int. J. Mol. Sci. 2020, 21, 6564 10 of 25

2.2.6. Impact of Pre-Analytical Sample Handling Procedures

Aliquots of individual EDTA-blood plasma samples from five different donors and a pooled plasmasample were subjected to magnetic bead immunoprecipitation with mAb 1E8. Aβ-peptides were elutedby heating in 20mM bicine, 0.6% 3-((3-cholamidopropyl) dimethylammonio)-1-propanesulfonate,pH 7.6 (bicine-CHAPS), as described previously [26], and diluted 4.8-fold with Diluent-35 [21].The diluted IP eluates were measured with the Aβ−3–40 assay freshly or after single or multiple freezingand re-thawing. We did not observe a substantial systematic impact of single or multiple freezing andthawing of the diluted IP eluates on the measured concentrations of Aβ−-3–40 (Figure 6a). A separateand independent control experiment did not suggest a general and systematic impact of a single orfive repeated freeze–thaw cycles on the measurement of Aβ38, Aβ40, and Aβ42 by the V-Plex Aβ

panel 1 (6E10) multiplex assay kit (Mesoscale Discovery, MSD) either (Figure S6). Consequently,pre-analytical freezing of IP eluates after dilution in Diluent-35 was implemented in the SOP. To assessthe impact of reducing the blood plasma volume and dilution with Diluent-35 prior to IP, three differentEDTA-plasma samples were studied. IPs were executed starting from 400 µL of EDTA plasma, from200 µL of EDTA plasma mixed with 200 µL of Diluent-35, and from 100 µL of EDTA-Plasma mixedwith 300 µL of Diluent-35. In each case, 100 µL of 5× triple detergent IP-buffer concentrate and 25 µLof 1E8 magnetic beads were added for overnight incubation. IP eluates were diluted 4.8-fold withDiluent-35 and stored at −80 ◦C until the analysis. As expected, after reducing the starting volumeof EDTA-plasma and diluting the samples with Diluent-35, lower concentrations of Aβ−3–40 weremeasured in the IP eluates. In two out of the three tested blood plasma samples, the measuredAβ−3–40 concentrations decreased in a linear fashion with sample dilution (Figure 6b). In the remainingsample, we observed a slight relative increase in Aβ−3–40 recovery after sample dilution, suggestingthe presence of interfering compounds (matrix effects) in undiluted EDTA-plasma.

2.2.7. Specificity of the Two-Step Immunoassay

To further evaluate the specificity of the two-step immunoassay for measuring Aβ−3–40 inhuman blood plasma, a number of control IPs were done and analyzed by the Aβ−3–40 immunoassay.The signals we obtained after magnetic bead immunoprecipitation from pooled EDTA-plasma usingthe anti-human Tau monoclonal HT7 and after 1E8-IP from 1% BSA in phosphate buffered saline(PBS) (“mock IP”) were below the detection range. The observations indicate absence of appreciablebackground signal owing to unspecific binding of plasma components to magnetic beads functionalizedwith an unrelated monoclonal antibody or from the 1E8-magnetic beads per se. Next, an Aβ-depletedplasma sample was prepared by two consecutive rounds of Aβ-IP from pooled plasma with a mixtureof magnetic beads coupled with anti Aβ monoclonals 6E10 or 4G8. The measured concentrations ofAβ−3–40 were 12.3 pg/mL in the Aβ-depleted sample and 239.3 pg/mL in the corresponding controlsample, indicating 95% reduction in Aβ−3–40 after Aβ depletion (Table 3).

Table 3. Specificity of the two-step immunoassay.

Sample Signal * Concentration (pg/mL) *

Zero calibrator (blank) 94 0HT7 magnetic bead IP (anti Tau) from plasma 87 Below detection range

1E8 magnetic bead IP from plasma 9140 239.41E8 magnetic bead IP from Aβ depleted plasma 150 12.34

1E8 mock IP from PBS/0.1% w/v BSA 86 Below detection range

* The indicated signals and calculated concentrations are the means of three technical replicates of the control IPeluates and two technical replicates (duplicate reads) of the zero calibrator (blank) on the assay plate. PBS: phoshatebuffered saline, BSA: bovine serum albumin.

Int. J. Mol. Sci. 2020, 21, 6564 11 of 25Int. J. Mol. Sci. 2020, 21, x FOR PEER REVIEW 11 of 26

Figure 6. Impact of pre-analytical sample handling procedures on the measurement of Aβ−3–40. (a)

Impact of pre-analytical freezing and thawing of diluted IP eluates. IP eluates obtained by 1E8

magnetic bead IP from aliquots of five individual and one pooled plasma sample were diluted with

Diluent-35 and measured in the Aβ−3–40 assay freshly or after one or five freeze–thaw cycles. Mean

calculated concentrations from two technical replicates on the assay plate and SDs are shown. (b)

Impact of changing the starting volume and diluting the blood plasma used for IP. IPs from three

different original blood plasma samples were done starting from 400 µL of EDTA plasma, from 200

µL of EDTA plasma mixed with 200 µL of Diluent-35, and from 100 µL of EDTA-Plasma plus 300 µL

of Diluent-35. The resulting IP eluates were diluted 4.8-fold with Diluent-35 and stored at −80 °C

until the analysis. Means ± SDs from triplicate reads are shown.

2.2.7. Specificity of the Two-Step Immunoassay

To further evaluate the specificity of the two-step immunoassay for measuring Aβ−3–40 in human

blood plasma, a number of control IPs were done and analyzed by the Aβ−3–40 immunoassay. The

signals we obtained after magnetic bead immunoprecipitation from pooled EDTA-plasma using the

anti-human Tau monoclonal HT7 and after 1E8-IP from 1% BSA in phosphate buffered saline (PBS)

(“mock IP”) were below the detection range. The observations indicate absence of appreciable

background signal owing to unspecific binding of plasma components to magnetic beads

functionalized with an unrelated monoclonal antibody or from the 1E8-magnetic beads per se. Next,

Figure 6. Impact of pre-analytical sample handling procedures on the measurement of Aβ−3–40.(a) Impact of pre-analytical freezing and thawing of diluted IP eluates. IP eluates obtained by 1E8magnetic bead IP from aliquots of five individual and one pooled plasma sample were diluted withDiluent-35 and measured in the Aβ−3–40 assay freshly or after one or five freeze–thaw cycles. Meancalculated concentrations from two technical replicates on the assay plate and SDs are shown. (b) Impactof changing the starting volume and diluting the blood plasma used for IP. IPs from three differentoriginal blood plasma samples were done starting from 400 µL of EDTA plasma, from 200 µL of EDTAplasma mixed with 200 µL of Diluent-35, and from 100 µL of EDTA-Plasma plus 300 µL of Diluent-35.The resulting IP eluates were diluted 4.8-fold with Diluent-35 and stored at −80 ◦C until the analysis.Means ± SDs from triplicate reads are shown.

2.2.8. Assay Interference

In an additional and separate control experiment, we assessed to what extent sAPPα, a relatedsynthetic model peptide (Aβ−23–16), and Aβ1–40 interfered with the measurement of Aβ−3–40 in a dilutedIP eluate from pooled human plasma. The model peptide Aβ−23–16 (APP649–687) (APP770 numbering)comprised 23 amino acids of the APP sequence N-terminal to the Aβ sequence followed by Aβ1–16.Both sAPPα and Aβ−23–16 are recognized by mAb101-1-1. The addition of sAPPα or Aβ−23–16 reducedthe measured Aβ−3–40 concentration in comparison with the control in a dose-dependent fashion(Table 4). In contrast, the addition of Aβ1–40 affected the measurement of Aβ−3–40 only marginally, even

Int. J. Mol. Sci. 2020, 21, 6564 12 of 25

at the highest tested concentration of 45 ng/mL. The experiment was repeated once with a comparableoutcome (data not shown). Notably, assay interference by sAPPα or related APP fragments is notexpected to represent a serious issue in the Aβ−3–40 two-step immunoassay, because the describedmagnetic bead IP with a covalently immobilized antibody against the Aβ N-terminus followed byheat elution in bicine-CHAPS pre-concentrates Aβ peptides very efficiently [21,26]. However, highconcentrations of sAPP or related APP derivatives might interfere with the correct direct measurementof absolute Aβ−3–40 concentrations in biological samples, dependent on the type of sample thatis analyzed.

Table 4. Analysis of assay interference.

Spike Measured Aβ−3–40 Concentration(pg/mL) * Recovery (% of Control)

Control (no spike) 183.93 100Aβ1–40, 4.5 ng/mL 185.72 101

Aβ1–40, 13.7 ng/mL 188.39 102Aβ1–40, 45 ng/mL 174.96 95sAPPα, 40 ng/mL 166.58 91

sAPPα, 120 ng/mL 154.51 84sAPPα, 363 ng/mL 136.16 74

Aβ−23–16, 4.5 ng/mL 162.23 88Aβ−23–16, 13.7 ng/mL 141.70 77Aβ−23–16, 45 ng/mL 88.18 48

* The indicated concentrations are the means of four (control) or two (spiked IP eluates) replicate measurements onthe assay plate.

2.2.9. Efficiency of the Aβ−3—40 Immunoprecipitation by 1E8-Magnetic Beads

To estimate how efficiently Aβ−3–40 is immunoprecipitated under the tested experimentalconditions, we performed two consecutive rounds of 1E8 IP from a pooled plasma sample. Comparisonof the measured Aβ−3–40 concentrations in the IP eluates of the first and second IP round indicatedhigh recovery after a single round of IP (Table 5). The measured Aβ−3–40 concentration after the secondIP was reduced to approximately 7.5% of that observed in the eluate obtained after the first IP.

Table 5. Efficiency of the 1E8 magnetic bead immunoprecipitation.

Sample 1E8-IP Round 1 1E8-IP Round 2

Signal * 8536 193Concentration (pg/mL) 229.77 17.29

* Mean of n = 3 technical replicates measured on the same assay plate.

2.2.10. Analytical Spike Recoveries

Two different individual EDTA plasma samples and a pooled plasma sample were spiked withsynthetic Aβ−3–40 at three different spike levels (10 pg, 20 pg, or 40 pg spiked into 400 µL of plasma)prior to immunoprecipitation with 1E8-magnetic beads. The diluted IP eluates obtained from neat andspiked samples were measured with the Aβ−3–40 assay. Spike recoveries were calculated by comparingthe amount of the spike added (in pg) with the amount detected in the spiked samples after subtractionof the endogenous Aβ−3–40 peptide found in the neat sample. The observed spike recoveries aresummarized in Table 6. Spike recoveries ranged from 32.5% to 99.6%, with mean values of 76% at thelow spike level (10 pg of Aβ−3–40 added to 400 µL of EDTA plasma), 94% at the intermediate spike level(20 pg of Aβ−3–40 added to 400 µL of EDTA plasma), and 67% at the high spike level (40 pg of Aβ−3–40

added to 400 µL of EDTA plasma). The findings suggest a certain degree of matrix interference, whichappears to vary between biological samples. In a control IP from 400 µL Diluent-35 spiked with 40 pgof Aβ−3–40, we observed 91.7% spike recovery.

Int. J. Mol. Sci. 2020, 21, 6564 13 of 25

Table 6. Analytical spike recoveries.

Measured Concentration (pg/mL)

Spike Level Neat Low Medium High

Sample 1 266.6 334.6 382.8 413.1Sample 4 243.1 265.0 381.4 468.5Plasma

pool 213.4 281.9 350.9 400.6

Control * n.a. n.d. n.d. 254.7

Spike Recovery (pg)

Spike Level Neat Low Medium High

Sample 1 n.a. 9.8 16.7 21.1Sample 4 n.a. 3.2 19.9 32.5Plasma

pool n.a. 9.9 19.8 27.0

Control * n.a. n.a. n.a. 36.7

Spike Recovery (%)

Spike Level Neat Low Medium High

Sample 1 n.a. 97.8 83.7 52.7Sample 4 n.a. 31.6 99.6 81.1Plasma

pool n.a. 98.6 99.0 67.4

Control * n.a. n.a. n.a. 91.7

* IP from 400 µL of Diluent-35 spiked with 40 pg of Aβ−3–40; n.a.: not applicable; n.d.: not determined.

2.2.11. Repeatability of Manual IPs from the Same Sample Performed in Parallel

Three different original EDTA plasma samples were studied. From each, threeimmunoprecipitations were executed in parallel, resulting in nine IP eluates in total. The observedcoefficient of variation between the parallel IP reactions in terms of measured Aβ−3—40 concentrationswas <10% (Table 7).

Table 7. Repeatability of manual IPs performed in parallel.

Aβ−3–40 pg/mL

IP 1 IP 2 IP 3 Mean SD %CV

Sample 1 237.45 211.09 233.63 227.39 14.24 6.26Sample 2 217.73 211.88 217.60 215.74 3.34 1.55Sample 3 183.42 207.80 203.71 198.31 13.06 6.58

Overall mean coefficient of variation (CV): 4.8%.

2.2.12. Day-to-Day Variance of the Manual IP Procedure

On four different days within 2 weeks, 1E8-magnetic bead IPs were executed from aliquots offive different human EDTA-plasma samples (termed A, B, C, D, E). The IP eluates obtained on each ofthe four days were diluted with Diluent-35 and stored frozen at −80 ◦C. Finally, all 20 samples werethawed and measured on a single assay plate. The coefficients of variation regarding the measuredAβ−3–40 concentrations in IP eluates prepared on four different days ranged from 3.6 to 20.1%, with amean value of 9.1% (Table S2).

2.2.13. Intra-Assay Variance/Repeatability of Aβ−3–40 Measurements in IP-Eluates

Diluted IP eluates obtained from three different original EDTA plasma samples were measured onthe same assay plate in eight technical replicates each. The observed coefficients of variation betweenthe replicate measurements were 2.2%, 2.7%, and 2%, respectively (Table S1).

Int. J. Mol. Sci. 2020, 21, 6564 14 of 25

2.2.14. Day-to-Day Variance of the Complete Two-Step Immunoassay and Normalization

To evaluate the intermediate imprecision (day-to-day variance) of the complete two-stepimmunoassay procedure (manual IP followed by immunoassay), we tested a set of three individualcontrol plasma samples and one control plasma pool in four different experiments. In each of thesefour assay runs, separate immunoprecipitations were performed from aliquots of the control samples.The measured Aβ−3–40 concentrations varied substantially between the four experiments, with a meancoefficient of variation of 29.6% (Table 8). Both the IP and the subsequent Aβ−3–40 immunoassayare expected to contribute to the overall intermediate imprecision (see above). In the next step, wenormalized the measured Aβ−3–40 concentrations by a QC-IP eluate included on each assay plate (seeabove). It served to calculate assay-plate correction factors for normalization of the error contributed bythe four independent runs of the Aβ−3–40 immunoassay only. The overall mean coefficient of variationof the complete two-step immunoassay after normalization was 16.2%. This was within the predefinedacceptance range (<20% CV), but still larger than the desired variance of <15% CV.

Table 8. Intermediate imprecision of the two-step immunoassay and normalization *.

Measured Concentrations (cm) Aβ−3–40(pg/mL)

QC-Sample Plate 1 Plate 2 Plate 3 Plate 4 Mean cm SD %CV

QC-1 185.0 150.2 221.0 323.0 219.8 74.6 33.9QC-2 258.0 160.2 257.3 336.0 252.9 72.0 28.5QC-3 180.8 145.4 242.0 298.4 216.7 67.6 31.2

QC-pool 227.2 192.7 234.2 336.4 247.6 61.9 25.0overall mean %CV 29.6

Normalized Concentrations (cnorm)(pg/mL)

QC-Sample Plate 1 Plate 2 Plate 3 Plate 4 Mean cnorm SD %CV

QC-1 208.5 167.7 235.7 252.4 216.1 37.0 17.1QC-2 290.8 178.9 274.5 262.2 251.7 49.9 19.8QC-3 203.7 162.3 258.2 233.2 214.4 41.2 19.2

QC-pool 256.1 215.0 249.8 246.0 246.0 21.3 8.7overall mean %CV 16.2

* Aβ−3–40 levels (cm) in three individual QC-plasma samples and one QC-plasma pool were determined in fourindependent runs of the two-step immunoassay (i.e., IP + immmunoassay). Normalized concentrations (cnorm)were calculated by means of specific correction factors for each assay plate.

2.3. Measurement of Aβ−3–40 in Patient Samples and Determination of Aβ Ratios as AD Biomarker Candidates

The relative levels of Aβ−3–40, Aβ40, and Aβ42 were measured in 40 human blood plasma samplesfrom patients with dementia of the Alzheimer’s type (AD-D, n = 23) and other dementias (OD,n = 17) by a semi-automated version of our two-step immunoassay. Here, the 1E8-magnetic bead IPswere executed on a CyBio FeliX liquid handling instrument (Analytik Jena GmbH, Jena, Germany)starting from 200 µL of EDTA-blood plasma mixed with 200 µL of H2O. The resulting IP eluateswere diluted 4.8-fold with Diluent-35 and stored in aliquots at −80 ◦C until they were measuredwith the novel Aβ−3–40 assay and the V-Plex Aβ panel 1 (6E10) multiplex assay kit. Importantly,the clinical cohort (n = 40 subjects) studied here is the same we investigated previously using ouroriginal two-step immunoassay procedure including Aβ pre-concentration by manual IP startingfrom 400 µL of EDTA-plasma [21]. In all 40 samples, Aβ−3–40, Aβ40, and Aβ42 measurements werein the detection range of the assay. Baseline statistics are summarized in Table S3. The measuredconcentrations of Aβ−3–40, Aβ40, and Aβ42, as well as the concentration ratios Aβ42/Aβ−3–40 andAβ42/Aβ40, showed lognormal distributions (Table S4). Thus, we performed pairwise comparisonsof means between the diagnostic groups OD (n = 17) and AD-D (n = 23) by multiple t-tests onlog2-transformed data. The p-values were corrected for multiple comparisons using the Holm–Sidak

Int. J. Mol. Sci. 2020, 21, 6564 15 of 25

method with α = 0.05 and without assuming a consistent SD. There were no statistically significantdifferences between the diagnostic groups OD and AD-D regarding the mean concentrations of Aβ−3–40

and Aβ40 (log2-transformed data). Aβ42, the Aβ42/Aβ40 ratio, and the Aβ42/Aβ−3–40 ratio (log2transformed data) were statistically significantly lower in the AD-D patients (Table 9).

Table 9. Comparison of means (log2 transformed data) between the diagnostic groups by multiplet-tests. AD-D, Alzheimer’s type dementia; OD, other dementias.

Aβ−3–40 * Aβ40 * Aβ42 * Aβ42/Aβ40 * Aβ42/Aβ−3–40 *

Significant? No No Yes Yes Yesp-value 0.65469 0.89737 0.01677 0.00003 0.00228

Mean of OD 6.022 6.684 3.658 −3.025 −2.364Mean of AD-D 6.061 6.694 3.483 −3.212 −2.577

Difference −0.0388 −0.0103 0.1756 0.1873 0.2134SE of difference 0.0860 0.0792 0.0702 0.0392 0.0652

t ratio 0.4508 0.1299 2.502 4.775 3.271df 38.00 38.00 38.00 38.00 38.00

Adjusted p-value ** 0.88076 0.89737 0.04947 0.00013 0.00910

* log2-transformed data; ** corrections for multiple comparisons were done using the Holm–Sidak method, withalpha = 0.05 without assuming a consistent SD. Number of t-tests: 5.

The diagnostic potential of plasma Aβ42 and the Aβ42/Aβ40 and Aβ42/Aβ−3–40 ratios for thediscrimination between OD versus AD-D in this clinical sample was compared by receiver operatingcharacteristic (ROC) analyses. The areas under the ROC curves (AUCs) were 0.760 for Aβ42, 0.790 forthe Aβ42/Aβ−3–40 ratio, and 0.854 for the Aβ42/Aβ40 ratio (Figure 7).

Int. J. Mol. Sci. 2020, 21, x FOR PEER REVIEW 16 of 26

* log2-transformed data; ** corrections for multiple comparisons were done using the Holm–Sidak

method, with alpha = 0.05 without assuming a consistent SD. Number of t-tests: 5.

The diagnostic potential of plasma Aβ42 and the Aβ42/Aβ40 and Aβ42/Aβ−3–40 ratios for the

discrimination between OD versus AD-D in this clinical sample was compared by receiver operating

characteristic (ROC) analyses. The areas under the ROC curves (AUCs) were 0.760 for Aβ42, 0.790 for

the Aβ42/Aβ−3–40 ratio, and 0.854 for the Aβ42/Aβ40 ratio (Figure 7).

Figure 7. Receiver operating characteristic (ROC) analysis of Aβ42, the Aβ42/Aβ40 ratio, and the

Aβ42/Aβ−3–40 ratio as discriminators of patients with other dementias (OD) and Alzheimer’s type

dementia (AD-D). The ROC curves of the indicated biomarker candidates in blood plasma are shown

in discriminating the diagnostic groups OD and AD-D. The areas under the curves (AUCs) were

0.760 (Aβ42), 0.790 (Aβ42/Aβ−3–40 ratio), and 0.854 (Aβ42/Aβ40 ratio), respectively.

3. Discussion

Affordable blood-based surrogate biomarker assays that can reliably detect AD associated

pathological changes at an early disease stage have enormous potential to support early and

preclinical diagnosis and the development of novel, disease modifying treatments. Recent studies

have shown that specific forms of the Aβ peptide can be detected in human blood plasma and

represent highly promising AD biomarker candidates. For example, a multimer detection system

employing two antibodies with overlapping N-terminal Aβ-epitopes for measuring Aβ

oligomerization after spiking synthetic Aβ into blood plasma samples has been developed [27].

Correlations between the plasma oligomerized Aβ and CSF Aβ42, amyloid-positron-emission

tomography (PET) results [28], and measures of cognitive functions [29,30] were reported. Other

research groups discovered that the plasma ratios Aβ42/Aβ40, Aβ1–40/Aβ1–42, and APP669–711/Aβ1-42 (i.e.,

Aβ−3–40/Aβ1–42), as measured by immunoprecipitation followed by mass spectrometry, predicted

brain amyloid pathology with high accuracy [19,20,23]. To the best of our knowledge, so far,

commercial antibodies against N-terminally elongated Aβ−3–X or immunoassays for the specific

measurement of Aβ−3–40 have not been available.

To set the groundwork for studying the biology of Aβ−3–40 in more detail in future studies and

for assessing Aβ−3–40 as a potential reference for accentuating the AD-associated selective decrease in

Aβ42 in blood plasma and possibly other biological fluids, we have characterized two novel

antibodies raised against the Aβ−3–X N-terminus and developed a highly selective sandwich

Figure 7. Receiver operating characteristic (ROC) analysis of Aβ42, the Aβ42/Aβ40 ratio, and the Aβ42/

Aβ−3–40 ratio as discriminators of patients with other dementias (OD) and Alzheimer’s type dementia (AD-D).The ROC curves of the indicated biomarker candidates in blood plasma are shown in discriminating thediagnostic groups OD and AD-D. The areas under the curves (AUCs) were 0.760 (Aβ42), 0.790 (Aβ42/Aβ−3–40

ratio), and 0.854 (Aβ42/Aβ40 ratio), respectively.

Int. J. Mol. Sci. 2020, 21, 6564 16 of 25

3. Discussion

Affordable blood-based surrogate biomarker assays that can reliably detect AD associatedpathological changes at an early disease stage have enormous potential to support early and preclinicaldiagnosis and the development of novel, disease modifying treatments. Recent studies have shown thatspecific forms of the Aβ peptide can be detected in human blood plasma and represent highly promisingAD biomarker candidates. For example, a multimer detection system employing two antibodies withoverlapping N-terminal Aβ-epitopes for measuring Aβ oligomerization after spiking synthetic Aβ intoblood plasma samples has been developed [27]. Correlations between the plasma oligomerized Aβ

and CSF Aβ42, amyloid-positron-emission tomography (PET) results [28], and measures of cognitivefunctions [29,30] were reported. Other research groups discovered that the plasma ratios Aβ42/Aβ40,Aβ1–40/Aβ1–42, and APP669–711/Aβ1-42 (i.e., Aβ−3–40/Aβ1–42), as measured by immunoprecipitationfollowed by mass spectrometry, predicted brain amyloid pathology with high accuracy [19,20,23].To the best of our knowledge, so far, commercial antibodies against N-terminally elongated Aβ−3–X orimmunoassays for the specific measurement of Aβ−3–40 have not been available.

To set the groundwork for studying the biology of Aβ−3–40 in more detail in future studies andfor assessing Aβ−3–40 as a potential reference for accentuating the AD-associated selective decrease inAβ42 in blood plasma and possibly other biological fluids, we have characterized two novel antibodiesraised against the Aβ−3–X N-terminus and developed a highly selective sandwich immunoassayfor measuring Aβ−3–40. Our observations show that the monoclonal antibodies 101-1-1 and 14-2-4can detect Aβ−3–40 without appreciable cross-reactivity with the canonical Aβ1–40 or N-terminallytruncated Aβ-variants in CIEF immunoassay and urea-SDS-PAGE/Western blot analysis. As it turnedout, neither of the two antibodies are specific for the free N-terminal valine. Both also recognize amyloidprecursor protein (APP), provided it does not carry the Swedish (K670N/M671L) double mutationlocated directly N-terminal to the β-secretase cleavage site. Whether or not the mAbs 101-1-1 and 14-2-4display preferences for monomeric, oligomeric, or aggregated forms of the N-terminally elongatedAβ−3–40 has not been assessed explicitly here. Published in vitro experiments including Thioflavin-Taggregation assay, size exclusion chromatography, and circular dichroism suggested a substantiallylower self-aggregation tendency of Aβ−3–40 than Aβ1–42 [23]. To the best of our knowledge, Aβ−3–40 hasnot been reported in amyloid plaques so far, and essentially no information regarding its potential rolein AD pathogenesis is available. In brain slices from APPLd transgenic mice analyzed in our presentstudy, mAb 101-1-1 appeared to bind neuronal cell bodies and to dystrophic neurites resemblingabnormal neuronal processes, but not to plaque cores. Presumably, the observed signals representcellular APP.

MAb 101-1-1 was selected to serve as the capture antibody in the finalized version of our novelAβ−3–40 immunoassay. The LLOD of the assay for measuring Aβ−3–40 was approximately 11 pg/mL, andthe quantitative assay range was approximately 22 pg/mL–7.5 ng/mL. We did not observe appreciablecross reactivity with Aβ1–40, Aβ1–42, Aβ−3–38, and sAPPα. In contrast, at high concentrations, syntheticAβ−3–42 produced measurable signals, indicating low cross reactivity at a level of <0.5%. The technicalintra-assay repeatability passed the predefined acceptance limit of <15% CV, and we did not findevidence for systematic spatial bias. The inter-assay CV of the calculated Aβ−3–40 concentrationin a QC IP eluate that was measured in five independent experiments within the assay validationcampaign was 16.3%, and thus outside of the predefined 15% CV acceptance range. It appeared thatthe large scatter was owing to a single experiment in which the calibrator dilution series had beenprepared in a slightly deviant way (to allow for determination of the ULOQ in this particular assayrun). Accordingly, we conclude that a strict SOP covering all pre-analytical and analytical steps indetail has to be implemented to minimize inter-assay variability.

The primary intended purpose of the novel assay is the assessment of Aβ−3–40 and the Aβ42/Aβ−3–40

ratio (or reverse) in human blood plasma as a candidate biomarker of AD. In view of publishedobservations from IP mass spectrometry studies [22–24], the blood concentration of Aβ−3–40 wasexpected to be small. Thus, we decided to pre-concentrate Aβs from plasma by magnetic bead

Int. J. Mol. Sci. 2020, 21, 6564 17 of 25

immunoprecipitation and to analyze the Aβ−3–40 plasma level by a two-step immunoassay procedure,essentially as described previously for the measurement of Aβ38, Aβ40, and Aβ42 [21]. A controlexperiment with a pooled blood plasma sample confirmed high yield of Aβ−3–40 after a single roundof 1E8 magnetic bead IP. Pre-analytical freezing and thawing of diluted IP eluates did not seem tosubstantially and systematically affect the measurement of Aβ−3–40. Consequently, freezing of thediluted IP eluates at −80 ◦C was implemented in our two-step immunoassay protocol to completelyseparate pre-analytical Aβ-enrichment from the actual measurement, and thus facilitate the executionof the two-step immunoassay under routine laboratory conditions. Furthermore, this allowed fortesting a semi-automated prototypic IP procedure executed at a different site (see below). We haveshown that the two-step immunoassay for Aβ−3–40 is highly selective. We did not observe appreciablesignals in plasma IP eluates from magnetic bead IP with an unrelated antibody or from a simulated IPwith 1E8 magnetic beads. Thus, unspecific binding of other plasma components to functionalized antimouse IgG magnetic beads or background signal directly stemming from the 1E8 coupled beads couldbe excluded. Depleting Aβ by two rounds of IP with anti Aβ mAbs 6E10 plus 4G8 prior to the 1E8 IPreduced the measured Aβ−3–40 concentration by approximately 95%.

Analytical spike recoveries in the two-step immunoassay varied substantially between the testedplasma samples and different spike levels indicating matrix interferences. Thus, the measured Aβ−3–40

levels in the diluted IP eluates do not allow for calculating the true, absolute concentrations of this Aβ

peptide in plasma samples, but have to be considered relative.The intermediate imprecision (between-run variance) of the complete manual two-step immunoassay

reached an acceptable CV of 16.2% only after normalization. Notably, in this case, the normalizationaddressed the technical error contributed by the Aβ−3–40 assay only, not the scatter of the manual IPprocedure. Without normalization, the mean between-run CV of the measured Aβ−3–40 concentrationsin four QC-samples was 29.6%, and thus clearly above the predefined acceptance criterion of <20%CV. In conclusion, appropriate QC samples allowing for normalization should be included in futureexperiments, whenever possible, and further technical improvement is desirable.

To finally test the novel Aβ−3–40 immunoassay on biological samples in a pilot study and todemonstrate the possibility of automation of the magnetic bead IP, we studied the same small clinicalsample we have previously described in detail [21]. In our current study, Aβ was pre-concentratedfrom 1:2 dilutions of blood plasma with water by a prototypic semi-automated 1E8 magnetic bead IPprotocol. The IP eluates were diluted 4.8-fold with Diluent-35 and frozen in aliquots for single useat −80 ◦C. For measuring Aβ−3–40, Aβ40, and Aβ42 at a different site, the samples were transportedon dry ice. We did not observe a statistically significant difference in the mean (log2 transformed)Aβ−3–40 levels between patients with AD-D (n = 23) and OD (n = 17). In contrast, Aβ42 and theAβ ratios Aβ42/Aβ−3–40 and Aβ42/Aβ40 were statistically significantly lower in patients with AD-Dcompared with those in the OD group. These findings are in agreement with the IP mass spectrometrydata collected in two independent cohorts by Nakamura et al. [23]. The areas under the ROC curvesin our small data set were 0.76 (Aβ42), 0.79 (Aβ42/Aβ−3–40 ratio), and 0.85 (Aβ42/Aβ40 ratio). Theseobservations support the idea that a suitable reference may serve to accentuate the pathologicalAD-associated decrease in soluble Aβ1–42 [22] (or Aβ42, respectively). Furthermore, our findingsshow that a prototypic semi-automated pre-analytical IP protocol works and that the adaptions andmodifications to the original manual two-step immunoassay that were implemented did not leadto masking or loss of the reduced Aβ42/Aβ40 plasma ratio as a peripheral diagnostic biomarker ofAD. Further studies employing independent and larger clinical cohorts will be required to assess thediagnostic potential of the two-step immunoassay for the Aβ42/Aβ−3–40 ratio in detail. Preferably,healthy controls and patients in different stages of AD should be included

In summary, we have developed a novel, highly selective sandwich immunoassay for measuringAβ−3–40 in biological samples. In combination with pre-analytical Aβ enrichment by magnetic beadIP, the assay can serve to measure the relative levels of Aβ−3–40 in human blood plasma. Thus, the

Int. J. Mol. Sci. 2020, 21, 6564 18 of 25

methodological groundwork has been set for future studies addressing the diagnostic potential of theAβ42/Aβ−3–40 ratio (or reverse) as a novel surrogate biomarker candidate of cerebral amyloid deposition.

4. Materials and Methods

Antibodies: The mouse monoclonal antibody (mAb) 101-1-1 was developed, produced, andpurified by Biogenes GmbH (Berlin, Germany) on a fee for service basis.·The synthetic peptideVKMDAEFRC-amide (Biosyntan, Berlin, Germany) was conjugated to bovine thyroglobulin as carrierprotein and served for immunization of mice. Hybridoma cells were screened by ELISA usingAβ1–40 and Aβ2–40 for negative screening and VKMDAEFRC-BSA-conjugate for positive screening.Clone 101-1-1 was obtained after two rounds of cloning. We received 88 mg of purified monoclonalantibody mAb 101-1-1 for antibody characterization, assay development, assay validation, andfurther use. Monoclonal anti Aβ−3–x, clone 14-2-4, was obtained from IBL International/Tecan,(Hamburg, Germany). MAb 1E8 (anti Aβ N-terminus) and mAb 5C3 (anti Aβ40) were purchasedfrom nanoTools GmbH (Teningen, Germany). MAb 280F2 (anti Aβ40) was obtained from SynapticSystems (Goettingen, Germany), mAb 6E10 from Biolegend (San Diego, CA, USA), anti APP mAb22C11 from Merck-Millipore (Darmstadt, Germany), and anti-Human Tau mAb HT7 from ThermoScientific (Waltham, MA USA). The rabbit monoclonal anti Aβ42 antibody mAb D3E10 was purchasedfrom Cell Signaling Technology (Danvers, MA, USA), while the rabbit polyclonal anti C-terminus APPA8717 was purchased from Sigma-Aldrich (Munich, Germany). SULFO-TAG Aβ40 detection antibodywas obtained from Mesoscale Discovery, Rockville, MD, USA. Biotinylated horse anti-mouse IgG waspurchased from Vector Laboratories (Burlingame, CA, USA), streptavidin-biotinylated peroxidasecomplex from GE Healthcare, Chicago, IL, USA), and goat anti mouse IgG peroxidase conjugatefrom Calbiochem/Millipore/Merck (Darmstadt, Germany). Biotinylated secondary anti-mouse andanti-rabbit antibodies, used in DAB-immunohistochemical staining, were obtained from DAKO/Agilent(Waldbronn, Germany). Fluorescent goat anti-Mouse IgG (DyLight 594) donkey anti-Rabbit IgG(DyLight 488) secondary antibodies were purchased from Thermo Scientific (Waltham, MA, USA).

Recombinant sAPPα was obtained from Sigma-Aldrich (Munich, Germany). Aβ−3–40, Aβ−3–38, andAβ−3–42 were synthesized as described in detail previously [25]. The synthetic peptides Aβ1–40, Aβ2–40,Aβ3–40, AβN3pE–40, Aβ4–40, and Aβ5–40 were obtained from AnaSpec (Fremont, CA, USA). The syntheticmodel peptide Aβ−23–16 (APP649–687) (H2N-GLTTRPGSGLTNIKTEEISEVKMDAEFRHDSGYEVHHQK-CONH2) was synthesized using standard solid-phase fluorenylmethoxycarbonyl (Fmoc) chemistry.Quality control of the purified product was performed by reversed-phase high-performance liquidchromatography coupled to a single-quadrupole mass spectrometer (Alliance/QDa, Waters Corporation,Milford, MA, USA) and by high-resolution mass spectrometry using a MALDI-TOF/TOF massspectrometer (UltrafleXtreme, Bruker, Billerica, MA, USA).

The CIEF immunoassay was performed as described before [8,26]. For chemiluminescencedetection, goat anti-mouse IgG, streptavidin-peroxidase conjugate, and Luminol/peroxide (all 3reagents obtained from ProteinSimple, San Jose, CA, USA) were employed.

Preparation of functionalized anti-Aβ magnetic beads: Dynabeads M280 sheep anti mouse IgG(ThermoFisher Scientific, Waltham, MA, USA) were functionalized by covalent coupling to mAb 1E8according to the manufacturer’s instructions and as described before [21]. Dynabeads M-270 Epoxy(Invitrogen/ThermoFisher Scientific, Waltham, MA, USA) were covalently coupled to mAb 101-1-1following the manufacturer’s instructions.

Antibody labeling: For small-scale biotinylation and SULFO-Tag labeling, 80 µL of a 1 mg/mLstock solution of mAb 101-1-1 or 280F2 was buffer-exchanged into MSD conjugation buffer (PBS, pH 7.9,preservative-free) on a Zeba Spin Desalting column 40 K MWCO, 0.5 mL (Thermo Scientific, Waltham,MA, USA). The buffer-exchanged antibody solution was divided into two halves, each containingapproximately 40 µg of IgG. For biotinylation, 3 µL of a 0.9 nmol/µL Sulfo-NHS-LC biotin solution(EZ-Link Micro Sulfo-NHS-LC Biotinylation kit, Thermo Scientific, Waltham, MA, USA) was added toapproximately 40 µg of IgG (challenge ratio: 10:1) and mixed immediately on a vortex mixer. After 2 h

Int. J. Mol. Sci. 2020, 21, 6564 19 of 25

incubation at 23 ◦C, excess biotin was removed by buffer exchange into MSD conjugate storage buffer(PBS, pH 7.4 containing 0.05% sodium azide) on a Zeba Spin Desalting column (40 K MWCO, 0.5 mL).For storage at −20 ◦C, glycerole was added to a final concentration of 50% (v/v). For SULFO-TAGlabeling, 1.8 µL of a 3 nmol/µL solution of MSD Gold SULFO-TAG NHS-Ester (Mesoscale Discovery,Rockville, MD, USA) was added dropwise to approximately 40 µg of buffer exchanged IgG (see above,challenge ratio: 20:1). After mixing by pipetting up and down, the reaction was incubated for 2 h atroom temperature in the dark. The remaining reagent was removed by buffer exchange into MSDconjugate storage buffer on a Zeba Spin desalting column (40 K MWCO, 0.5 mL). Storage: at 4 ◦C inthe dark.

For the biotinylation of mAb 101-1-1 on a larger scale, 159 µL of 101-1-1 [6.3 mg/mL] was mixedwith 841 µL of PBS (0.1 M sodium phosphate/0.15 M NaCl, pH 7.2, Thermo Scientific, Waltham,MA, USA) and buffer exchanged into PBS (0.1 M sodium phosphate/0.15 M NaCl, pH 7.2, ThermoScientific) on a Zeba Spin desalting column (5 mL, 7000 MWCO, Thermo Scientific, Waltham, MA, USA).Then, 7.4 µL of a 9 nmol/µL Sulfo-NHS-LC biotin stock solution was added (challenge ratio 10:1)and mixed on a vortex immediately. After incubation for 2 h at 23 ◦C, excess biotin was removedby buffer exchange into PBS. The protein concentration was measured by microplate BCA proteinassay (PIERCE/ThermoFisher Scientific, Waltham, MA, USA) and the level of biotin incorporation wasdetermined by HABA (4-hydroxyazobenzene-2-carboxylic acid) assay in a microplate according to theinstructions to the EZ-Link Sulfo-NHS-Biotinylation Kit (Themo Scientific, Waltham, MA, USA). BSA andsodium azide were added to the biotinylated antibody to final concentrations of 1% and 0.05%, respectively.For long term storage, half of the material was aliquoted and frozen at −80 ◦C. The remaining biotinylatedantibody stock was supplemented with 40% (v/v) ethylene glycol and stored in aliquots at −20 ◦C.

Urea-SDS-PAGE and semi dry blotting: Aβ-peptides were separated by urea-bicine/bis-tris/tris/sulfate SDS-polyacrylamide gel electrophoresis [31] on 12%T/5% C gels and subsequently transferredon PVDF membranes by semi dry blotting, essentially as described previously [32,33]. The proteintransfer occurred at a constant current of 0.8 mA/cm2 for 30 min with a voltage limit of maximum 30 V.

Bis-Tris Gradient Gel electrophoresis and Western blotting: Proteins and peptides were separatedon a 4–12% Bis-Tris gradient gel (Anamed Elektrophorese GmbH, Groß-Bieberau, Germany) at 200 Vfor 30 min with MES (2-(N-morpholino)ethanesulfonic acid) running buffer (50 mM MES, 50 mMTris, 0.1% w/v SDS, 1 mM EDTA). NuPAGE antioxidant (Thermo Fisher Sientific, Waltham, MA,USA) was added to the upper buffer chamber. The proteins were subsequently blotted onto PVDFmembranes using 25 mM Bicine, 25 mM Bis-Tris pH 7.2, 1 mM EDTA, and 10% (v/v) methanol transferbuffer with NuPAGE antioxidant at 20 V for 60 min (MiniGel Tank and Blot Module obtained fromInvitrogen/Thermoscientific, Waltham, MA, USA).

Immunohistochemistry: Experiments involving animal tissues were approved by the local animalcare and use committee (Landesamt für Verbraucherschutz und Lebensmittelsicherheit (LAVES), LowerSaxony). Brain tissue slides from 12-month-old 5XFAD [34] and 15-month-old APP/Ld transgenicmice [35] were stained as published elsewhere [36]. In brief, 4 µm paraffin sections were deparaffinizedin xylene and rehydrated in a series of ethanol, followed by incubation in PBS containing 1% H2O2 toblock endogenous peroxidases. Antigen-retrieval was carried using microwave treatment in 0.01 Mcitrate buffer pH 6.0 and incubation in 88% formic acid. Non-specific binding sites were blocked with4% skim milk in PBS containing 10% fetal calf serum. This was followed by overnight incubationwith 101-1-1 (1:10,000) or APP C-term (A8717, Sigma Aldrich, Munich, Germany, 1:3000) in a humidchamber. Staining was visualized using the ABC method with a Vectastain Kit (Vector Laboratories,Burlingame, CA, USA) and diaminobenzidine. In the case of fluorescent staining, 101-1-1 (1:6000) andthe Aβ42-specific antibody D3E10 (1:1000) (Cell Signaling Danvers, MA, USA) were used and detectedusing anti-mouse-DyLight-594 or anti-rabbit-Dylight-488 antibodies (both Thermo Fisher, Waltham,MA, USA).

Manual and automated immunoprecipitations: Magnetic bead immunoprecipitations (IPs) byhand starting from 400 µL of EDTA-blood plasma were performed as described previously [21].

Int. J. Mol. Sci. 2020, 21, 6564 20 of 25

An adapted protocol for a semi-automated IP procedure was executed on a CyBio FeliX liquid handlingInstrument (Analytik Jena, Jena, Germany) equipped with BioShake 3000-Telm (Q Instruments, Jena,Germany) and MAGNUM FLX Enhanced Universal Magnet Plate, Alpaqua Engineering, LLC, Beverly,MA, USA.

Aliquots of EDTA-blood plasma samples (approximately 500 µL per sample), which had beenstored at −80 ◦C in Matrix vials, were thawed and transferred into 1.5 mL reaction vials. The vials weremixed vigorously on a vortex mixer (5 × 10 s), and insoluble material was pelleted by centrifugationfor 10 min at 10,000× g at room temperature in a fixed angle rotor. Then, 220 µL of each supernatantwas transferred into a 96-well sample plate (Deep Well MegaBlock®, 96 wells, 2.2 mL, PP (Sarstedt,Nümbrecht, Germany)) and manually mixed with 220 µL of H2O.

The sample plate was mounted into the CyBio FeliX instrument and preparation of immunoprecipitationmixes was carried out automatically by adding 100 µL of 5× IP buffer concentrate [21] and 25 µL of 1E8magnetic beads to 400 µL of the diluted samples. Finally, the instrument transferred the preparations into aprocess plate (Deep Well MegaBlock®, 96 wells, 2.2 mL, PP (Sarstedt, Nümbrecht, Germany)).

The process plate was then manually placed in the “process” position of the CyBio FeliX instrumentand the immunoprecipitation process was executed automatically. The process included an 18-hincubation at room temperature, where the samples were regularly mixed. Then, the process plate wasmoved into a magnet position for 120 s for immobilization of the magnetic beads. The supernatant wasaspirated and discarded, and the process plate was moved back to the process position. Per well, 1 mLof wash buffer (PBS containing 0.1% w/v BSA) was added and the beads were washed for 5 min at roomtemperature with agitation. The plate was moved to the magnet position for bead immobilizationand removal of the wash fluid as described. Then, the plate was moved back to the process positionfor a second wash. In total, the magnetic bead immune complexes were washed 3× for 5 min withPBS/0.1% w/v BSA and once for 3 min with 1 mL of 10 mM Tris-HCl, pH 7.5. After removal of the washbuffer, 2 × 25 µL of elution buffer (20 mM Bicine, pH 7.6, 0.6% w/v CHAPS) were added, and the beadswere resuspended by aspirating and dispensing. The resuspended beads were then transferred into apreheated elution plate (Deep Well MegaBlock®, 96 wells, 1.2 mL PP (Sarstedt, Nümbrecht, Germany))positioned on the BioShake and incubated for 5 min at 95 ◦C and 800 rpm to allow elution of thebound Aβ peptides. Then, the suspension was quantitatively moved to a 96-well plate (Deepwell plate96/500 µL Protein LoBind (Eppendorf, Hamburg, Germany)) located on the magnet and incubated for120 s for cooling and bead immobilization. Subsequently, 190 µL of Diluent-35 was added to each well(4.8-fold dilution of the IP eluates) and the samples were mixed. Finally, 200 µL of each diluted IPeluate was transferred into a separate plate (Deepwell plate 96/500 µL Protein LoBind (Eppendorf,Hamburg, Germany)). Aliquots of 60 µL each were manually pipetted into 1.5 mL reaction vials andfrozen at −80 ◦C.

Measurements of Aβ40 and Aβ42: Aβ40 and Aβ42 levels in 4.8-fold diluted IP eluates were measuredwith the V-Plex Aβ panel 1 (6E10) multiplex assay kit (Mesoscale Discovery, MSD, Rockville, MD, USA)according to the manufacturer’s instructions and as described previously [21]. In the current version ofthe two-step immunoassay protocol, the 4.8-fold diluted IP eluates were measured after storage at −80 ◦Cand singular thawing.

Development of the Aβ−3–40 immunoassay: The novel sandwich immunoassay for the measurementof Aβ−3–40 was developed on the MSD technology platform (Mesoscale Discovery, MSD, Rockville, MD,USA) using MSD Gold 96-well Small Spot streptavidin plates and various MSD reagents, including 20×wash buffer, Diluent-100, Diluent-35, Blocker A, 4×Read buffer, and SULFO-TAG Aβ40 detection antibody.In the final assay protocol (SOP), 150 µL of 3% (w/v) Blocker-A in wash buffer was added to the wells ofa 96-well Small Spot streptavidin plate. The plate was sealed with an adhesive foil and incubated for1 h at room temperature with shaking (500 rpm) on an Eppendorf Thermomix with a lightproof cover.The blocking solution was discarded and the plate was washed three times with 200 µL of wash buffer perwell. For plate coating, 25 µL of a 2 µg/mL dilution of the biotinylated capture antibody in Diluent-100was added per well. The plate was sealed with an adhesive foil and incubated for 1 h at room temperature

Int. J. Mol. Sci. 2020, 21, 6564 21 of 25

with shaking (as described above). After three washes with 200 µL of wash buffer per well, 25 µL ofsample or calibrator dilution per well was pipetted into the assay plate. The plate was sealed with anadhesive foil and incubated for 1 h at room temperature with shaking (see above). After three washeswith 200 µL of wash buffer per well, 25 µL of detection antibody diluted in Diluent-100 was added to eachwell. Again, the plate was sealed with an adhesive foil and incubated for 1 h at room temperature withshaking (500 rpm) on an Eppendorf Thermomix with a lightproof cover. After three washes with 200 µLof wash buffer per well, 150 µL of 2× Read buffer was added to each well by reverse pipetting. Formationof bubbles must be avoided at this stage. The electro-chemiluminescent signals were recorded on an MSDQuickplex SQ120 reader and analyzed with the MSD Discovery Workbench Software version 4.0.12.

The synthetic calibrator peptide Aβ−3–40 was synthesized as described before [25]. The lyophilizedpeptide was initially solubilized in DMSO (dimethyl sulfoxide) at a concentration of 1 mg/mL andstored in aliquots at −80 ◦C. Starting from this master stock solution in DMSO, a 600 ng/mL calibratorstock solution for use in the novel Aβ−3–40 immunoassay was prepared in Diluent-35 and stored inaliquots for single use at −80 ◦C. In most experiments, seven serial calibrator dilutions in Diluent-35plus a zero calibrator (blank) were measured for calculating the standard curve by four-parameterlogistic curve fitting. The seven calibrator solutions (standards) were prepared by serial fourfolddilution in Diluent-35 starting with the highest standard to be measured. During the assay validationcampaign, in most cases, the highest standard had a concentration of 1875 pg/mL. For the study of theclinical sample, the highest standard contained 7.5 ng/mL of Aβ−3–40. In one experiment, an extendedfourfold calibrator dilution series including ten fourfold serial dilutions starting at 120 ng/mL pluszero calibrator (blank) was analyzed to determine the ULOQ. The alternative synthetic D7-Calibrator(VKMDAEFRH followed by 5× PEG2 (8-amino-3,6-dioxaoctanoic acid) and the Aβ40 C-terminusGLMVGGVV) (Mr: 2571.4)) was synthesized by Biosyntan (Berlin, Germany).

Cell culture: H4 cells stably overexpressing Swedish mutant APP751SWE [37] were cultured inD-MEM/F12, supplemented with 10% fetal calf serum, 2 mM L-Alanyl-L-Glutamine, non-essentialamino acids, and 500 µg/mL Hygromycin. To generate H4 cells stably overexpressing wild typeAPP751, the human amyloid beta A4 protein isoform b precursor cDNA (also known as PreA4751, APP 751) was subcloned into a pCI-neo mammalian expression vector (Promega, Walldorf,Germany). H4 neuroglioma cells from the exponential growth phase were seeded at 5 × 105

cells/cm2 in H4 growth medium (DMEM supplemented with 4.5 g/L glucose, 4 mM L-glutamine,10% (v/v) fetal bovine serum, 100 IU/mL penicillin, and 100µg/mL streptomycin) 24 h before thetransfection. Medium changes with serum free DMEM/Ham´s F12 with G5 supplement wereperformed 3 h before and directly before the transfection. Plasmid DNA and CaCl2 were mixedwith N,N-Bis(2-hydroxyethyl)-2-aminoethanesulfonic acid (BES) buffer with pH 6.98, yielding finalconcentrations of 25 mM BES, 140 mM NaCl, 0.75 mM Na2HPO4, 125 mM CaCl2, and 37.5 ng/µLplasmid DNA. For the transfection of a growth area of 9.5 cm2, 200 µL of the mixture was incubatedfor 20 min at room temperature in the dark and subsequently added to the cells. The culture mediumwas changed to growth medium 24 h after transfection. After 48 h, 500µg/mL G418 (Thermo FisherScientific, Waltham, MA, USA) was added and resistant clones were selected by limiting dilution at<0.2 cells/well in H4 growth medium with 500µg/mL G418. The cells were maintained in the presenceof 500 µg/mL G418. Six clones were isolated and assessed for the level of APP expression. The cellswere cultured in DMEM GlutaMAX (Thermo Fisher, Waltham, MA, USA) supplemented with 10%superior fetal bovine serum (Biochrom, Berlin, Germany) and 500 µg/mL G418.

Study approval and study cohort. The study was conducted in accordance with the Declaration ofHelsinki. The pseudonomized collection of biological samples and clinical data in a local biobank andtheir use in biomarker studies was approved by the ethics committee of the University of Göttingen(9/2/16). All subjects or their legal representatives gave their informed consent prior to inclusion.The clinical sample comprising n = 23 patients with dementia of the Alzheimer’s type (AD-D) andn = 17 patients with dementia due to other reasons (OD) has been described in detail in a previous

Int. J. Mol. Sci. 2020, 21, 6564 22 of 25

study [21]. The preparation and storage of blood plasma samples according to standardized protocolwere described before (ibid.).

Statistical analysis: For statistical analysis and graphical representation of data, we employedGraphPad Prism 8.4 and IBM SPSS Statistics 26.

Normalization between assay plates. On each assay plate in the core validation campaign, weincluded the measurement of an aliquot of a quality control (QC) IP eluate in four technical replicates.It served to assess the between-run variance of the Aβ−3–40 assay and to allow for testing a methodfor normalization between assay plates. The QC-IP eluate was prepared by running several IPsin parallel from pooled human blood plasma and combining the obtained IP eluates after 4.8-folddilution to a single QC-IP eluate. Aliquots for single use were stored frozen at −80 ◦C until the analysis.For normalization, the average concentration of Aβ−3–40 in the QC-IP eluate (mean of four independentassay runs) was calculated as follows:

cavg =n∑

i=1

cm (1)

where cm = measured concentration, mean of the replicate measurements on the respective assay plate,and cavg = mean concentration (average of n = 4 plates considered).

Subsequently, plate correction factors (k) for each assay plate were calculated as follows:

k =cavg

cm(2)

Finally, the normalized Aβ−3–40 concentrations (cnorm) were calculated with the following formula:

cnorm = cm ∗ k (3)

The normalization by the QC IP eluate was applied to the data obtained with aliquots of a set ofthree individual control plasma samples and a control plasma pool that were analyzed by the completetwo-step immunoassay procedure (i.e., IP followed by Aβ−3–40 measurement) in four independentexperiments. It should be noted that the described normalization procedure considered only theexperimental error contributed by the 96-well plate immunoassay, but not the error contributed by themanual immunoprecipitation.

Supplementary Materials: Supplementary materials can be found at http://www.mdpi.com/1422-0067/21/18/6564/s1.

Author Contributions: Conceptualization, H.W.K. and J.W.; Formal analysis, H.W.K.; Investigation, P.R., S.Z.,D.O., and L.R.; Methodology, H.W.K., P.R., A.M., and O.J.; Resources, A.M., O.W., J.V., T.J.O., O.J., I.B., I.L., andH.-J.K.; Supervision J.W.; Validation, H.W.K. and P.R.; Visualization, H.W.K. and O.W.; Writing—original draft,H.W.K. and O.W.; Writing—review & editing, H.W.K., S.Z., O.W., J.V., D.O., T.J. Oberstein, O.J., I.L., and H.-J.K.All authors have read and agreed to the published version of the manuscript.

Funding: J.W. is supported by an Ilídio Pinho professorship and iBiMED (UID/BIM/04501/2013) at the University ofAveiro, Portugal. J.W. is also supported by FCT project PTDC/DTP_PIC/5587/2014 and POCI-01-0145-FEDER-016904. J.V.was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation)—project number 413501650.

Acknowledgments: The authors thank A. Jahn-Brodmann, T. Liepold, and L. van Werven for expert technical assistance.

Conflicts of Interest: J.W. is a consultant or member of the advisory board for Abbott, Biogen, BoehringerIngelheim, Immungenetics, Lilly, MSD Sharp & Dohme, and Roche Pharma. He received fees for lectures fromActelion Pharmaceuticals, Amgen, Janssen-Cilag, and Pfizer. A.M. is a full-time employee of IBL InternationalGmbH, Tecan Group Company, Hamburg, Germany. D.O. and I.L. are full time employees of Roboscreen GmbH,Leipzig, Germany.

Abbreviations

AD Alzheimer’s diseaseAβ amyloid β

APP amyloid precursor protein

Int. J. Mol. Sci. 2020, 21, 6564 23 of 25

PET positron-emission tomographymAb monoclonal antibodyBSA bovine serum albuminIP immunoprecipitationCHAPS 3-(3-cholamidopropyl) dimethylammonio)-1-propanesulfonateLLOD lower limit of detectionLLOQ lower limit of quantificationHABA 4-hydroxyazobenzene-2-carboxylic acidBCA bicinchoninic acidSOP standard operating procedureULOQ upper limit of quantificationCV coefficient of variationPEG2 8-Amino-3,6-dioxaoctanoic acidQC quality controlAD-D dementia of the Alzheimer’s typeOD other dementiasROC receiver operating characteristicAUC area under the curveBES N,N-Bis(2-hydroxyethyl)-2-aminoethanesulfonic acid

References

1. Lane, C.A.; Hardy, J.; Schott, J.M. Alzheimer’s disease. Eur. J. Neurol. 2018, 25, 59–70. [CrossRef]2. Glenner, G.G.; Wong, C.W. Alzheimer’s disease: Initial report of the purification and characterization of a

novel cerebrovascular amyloid protein. Biochem. Biophys. Res. Commun. 1984, 120, 885–890. [CrossRef]3. Masters, C.L.; Simms, G.; Weinman, N.A.; Multhaup, G.; McDonald, B.L.; Beyreuther, K. Amyloid plaque

core protein in Alzheimer disease and Down syndrome. Proc. Natl. Acad. Sci. USA 1985, 82, 4245–4249.[CrossRef] [PubMed]

4. Selkoe, D.J. The cell biology of beta-amyloid precursor protein and presenilin in Alzheimer’s disease.Trends Cel Biol. 1998, 8, 447–453. [CrossRef]

5. Huang, Y.R.; Liu, R.T. The Toxicity and Polymorphism of beta-Amyloid Oligomers. Int. J. Mol. Sci. 2020, 21,4477. [CrossRef]

6. Portelius, E.; Bogdanovic, N.; Gustavsson, M.K.; Volkmann, I.; Brinkmalm, G.; Zetterberg, H.; Winblad, B.;Blennow, K. Mass spectrometric characterization of brain amyloid beta isoform signatures in familial andsporadic Alzheimer’s disease. Acta Neuropathol. 2010, 120, 185–193. [CrossRef]

7. Wildburger, N.C.; Esparza, T.J.; LeDuc, R.D.; Fellers, R.T.; Thomas, P.M.; Cairns, N.J.; Kelleher, N.L.;Bateman, R.J.; Brody, D.L. Diversity of Amyloid-beta Proteoforms in the Alzheimer’s Disease Brain. Sci. Rep.2017, 7, 9520. [CrossRef]

8. Zampar, S.; Klafki, H.W.; Sritharen, K.; Bayer, T.A.; Wiltfang, J.; Rostagno, A.; Ghiso, J.; Miles, L.A.; Wirths, O.N-terminal heterogeneity of parenchymal and vascular amyloid-beta deposits in Alzheimer’s disease.Neuropathol. Appl. Neurobiol. 2020, 10. [CrossRef]

9. Wirths, O.; Zampar, S. Emerging roles of N- and C-terminally truncated Abeta species in Alzheimer’s disease.Expert Opin. Ther. Targets 2019, 23, 991–1004. [CrossRef]

10. Dunys, J.; Valverde, A.; Checler, F. Are N- and C-terminally truncated Abeta species key pathological triggersin Alzheimer’s disease? J. Biol. Chem. 2018, 293, 15419–15428. [CrossRef]

11. Saido, T.C.; Iwatsubo, T.; Mann, D.M.; Shimada, H.; Ihara, Y.; Kawashima, S. Dominant and differentialdeposition of distinct beta-amyloid peptide species, A beta N3(pE), in senile plaques. Neuron 1995, 14,457–466. [CrossRef]

12. Schlenzig, D.; Ronicke, R.; Cynis, H.; Ludwig, H.H.; Scheel, E.; Reymann, K.; Saido, T.; Hause, G.; Schilling, S.;Demuth, H.U. N-Terminal pyroglutamate formation of Abeta38 and Abeta40 enforces oligomer formation andpotency to disrupt hippocampal long-term potentiation. J. Neurochem. 2012, 121, 774–784. [CrossRef] [PubMed]

13. He, W.; Barrow, C.J. The A beta 3-pyroglutamyl and 11-pyroglutamyl peptides found in senile plaque havegreater beta-sheet forming and aggregation propensities in vitro than full-length A beta. Biochemistry 1999,38, 10871–10877. [CrossRef] [PubMed]

Int. J. Mol. Sci. 2020, 21, 6564 24 of 25

14. Gunn, A.P.; Wong, B.X.; Johanssen, T.; Griffith, J.C.; Masters, C.L.; Bush, A.I.; Barnham, K.J.; Duce, J.A.;Cherny, R.A. Amyloid-beta Peptide Abeta3pE-42 Induces Lipid Peroxidation, Membrane Permeabilization,and Calcium Influx in Neurons. J. Biol. Chem. 2016, 291, 6134–6145. [CrossRef] [PubMed]

15. Russo, C.; Violani, E.; Salis, S.; Venezia, V.; Dolcini, V.; Damonte, G.; Benatti, U.; D’Arrigo, C.; Patrone, E.;Carlo, P.; et al. Pyroglutamate-modified amyloid beta-peptides–AbetaN3(pE)–strongly affect cultured neuronand astrocyte survival. J. Neurochem. 2002, 82, 1480–1489. [CrossRef] [PubMed]

16. Scheidt, H.A.; Das, A.; Korn, A.; Krueger, M.; Maiti, S.; Huster, D. Structural characteristics of oligomers formed bypyroglutamate-modified amyloid beta peptides studied by solid-state NMR. Phys. Chem. Chem. Phys. 2020, 22,16887–16895. [CrossRef]

17. Blennow, K.; Zetterberg, H. The Past and the Future of Alzheimer’s Disease Fluid Biomarkers. J. Alzheimer’s Dis.2018, 62, 1125–1140. [CrossRef]

18. Cummings, J. The Role of Biomarkers in Alzheimer’s Disease Drug Development. Adv. Exp. Med. Biol. 2019,1118, 29–61. [CrossRef]

19. Ovod, V.; Ramsey, K.N.; Mawuenyega, K.G.; Bollinger, J.G.; Hicks, T.; Schneider, T.; Sullivan, M.; Paumier, K.;Holtzman, D.M.; Morris, J.C.; et al. Amyloid beta concentrations and stable isotope labeling kinetics of humanplasma specific to central nervous system amyloidosis. Alzheimer’s Dement. 2017, 13, 841–849. [CrossRef]

20. Schindler, S.E.; Bollinger, J.G.; Ovod, V.; Mawuenyega, K.G.; Li, Y.; Gordon, B.A.; Holtzman, D.M.; Morris, J.C.;Benzinger, T.L.S.; Xiong, C.; et al. High-precision plasma beta-amyloid 42/40 predicts current and futurebrain amyloidosis. Neurology 2019, 93, e1647–e1659. [CrossRef]

21. Shahpasand-Kroner, H.; Klafki, H.W.; Bauer, C.; Schuchhardt, J.; Huttenrauch, M.; Stazi, M.; Bouter, C.;Wirths, O.; Vogelgsang, J.; Wiltfang, J. A two-step immunoassay for the simultaneous assessment of Abeta38,Abeta40 and Abeta42 in human blood plasma supports the Abeta42/Abeta40 ratio as a promising biomarkercandidate of Alzheimer’s disease. Alzheimer’s Res. Ther. 2018, 10, 121. [CrossRef] [PubMed]

22. Kaneko, N.; Nakamura, A.; Washimi, Y.; Kato, T.; Sakurai, T.; Arahata, Y.; Bundo, M.; Takeda, A.; Niida, S.; Ito, K.;et al. Novel plasma biomarker surrogating cerebral amyloid deposition. Proc. Jpn. Acad. Ser. B Phys. Biol. Sci. 2014,90, 353–364. [CrossRef] [PubMed]

23. Nakamura, A.; Kaneko, N.; Villemagne, V.L.; Kato, T.; Doecke, J.; Dore, V.; Fowler, C.; Li, Q.X.; Martins, R.;Rowe, C.; et al. High performance plasma amyloid-beta biomarkers for Alzheimer’s disease. Nature 2018,554, 249–254. [CrossRef]

24. Kaneko, N.; Yamamoto, R.; Sato, T.A.; Tanaka, K. Identification and quantification of amyloid beta-relatedpeptides in human plasma using matrix-assisted laser desorption/ionization time-of-flight mass spectrometry.Proc. Jpn. Acad. Ser. B Phys. Biol. Sci. 2014, 90, 104–117. [CrossRef] [PubMed]

25. Beyer, I.; Rezaei-Ghaleh, N.; Klafki, H.W.; Jahn, O.; Haussmann, U.; Wiltfang, J.; Zweckstetter, M.; Knölker, H.J.Solid-Phase Synthesis and Characterization of N-Terminally Elongated Abeta-3-x -Peptides. Chemistry 2016,22, 8685–8693. [CrossRef] [PubMed]

26. Haussmann, U.; Jahn, O.; Linning, P.; Janssen, C.; Liepold, T.; Portelius, E.; Zetterberg, H.; Bauer, C.;Schuchhardt, J.; Knölker, H.J.; et al. Analysis of amino-terminal variants of amyloid-beta peptides bycapillary isoelectric focusing immunoassay. Anal. Chem. 2013, 85, 8142–8149. [CrossRef]

27. An, S.S.A.; Lee, B.S.; Yu, J.S.; Lim, K.; Kim, G.J.; Lee, R.; Kim, S.; Kang, S.; Park, Y.H.; Wang, M.J.; et al. Dynamicchanges of oligomeric amyloid beta levels in plasma induced by spiked synthetic Abeta42. Alzheimer’s Res. Ther.2017, 9, 86. [CrossRef]

28. Wang, M.J.; Yi, S.; Han, J.Y.; Park, S.Y.; Jang, J.W.; Chun, I.K.; Kim, S.E.; Lee, B.S.; Kim, G.J.; Yu, J.S.; et al.Oligomeric forms of amyloid-beta protein in plasma as a potential blood-based biomarker for Alzheimer’sdisease. Alzheimer’s Res. Ther. 2017, 9, 98. [CrossRef]

29. Meng, X.; Li, T.; Wang, X.; Lv, X.; Sun, Z.; Zhang, J.; Su, F.; Kang, S.; Kim, S.; An, S.S.A.; et al. Associationbetween increased levels of amyloid-beta oligomers in plasma and episodic memory loss in Alzheimer’sdisease. Alzheimer’s Res. Ther. 2019, 11, 89. [CrossRef]

30. Lee, J.J.; Choi, Y.; Chung, S.; Yoon, D.H.; Choi, S.H.; Kang, S.M.; Seo, D.; Park, K.I. Association of PlasmaOligomerized Beta Amyloid with Neurocognitive Battery Using Korean Version of Consortium to Establisha Registry for Alzheimer’s Disease in Health Screening Population. Diagnostics 2020, 10, 237. [CrossRef]

31. Klafki, H.W.; Wiltfang, J.; Staufenbiel, M. Electrophoretic separation of betaA4 peptides (1-40) and (1-42).Anal. Biochem. 1996, 237, 24–29. [CrossRef] [PubMed]

Int. J. Mol. Sci. 2020, 21, 6564 25 of 25

32. Wiltfang, J.; Smirnov, A.; Schnierstein, B.; Kelemen, G.; Matthies, U.; Klafki, H.W.; Staufenbiel, M.; Huther, G.;Ruther, E.; Kornhuber, J. Improved electrophoretic separation and immunoblotting of beta-amyloid (A beta)peptides 1-40, 1-42, and 1-43. Electrophoresis 1997, 18, 527–532. [CrossRef] [PubMed]

33. Schieb, H.; Kratzin, H.; Jahn, O.; Mobius, W.; Rabe, S.; Staufenbiel, M.; Wiltfang, J.; Klafki, H.W. Beta-amyloidpeptide variants in brains and cerebrospinal fluid from amyloid precursor protein (APP) transgenic mice:Comparison with human Alzheimer amyloid. J. Biol. Chem. 2011, 286, 33747–33758. [CrossRef]

34. Oakley, H.; Cole, S.L.; Logan, S.; Maus, E.; Shao, P.; Craft, J.; Guillozet-Bongaarts, A.; Ohno, M.; Disterhoft, J.;Van Eldik, L.; et al. Intraneuronal beta-amyloid aggregates, neurodegeneration, and neuron loss in transgenicmice with five familial Alzheimer’s disease mutations: Potential factors in amyloid plaque formation.J. Neurosci. 2006, 26, 10129–10140. [CrossRef]

35. Van Dorpe, J.; Smeijers, L.; Dewachter, I.; Nuyens, D.; Spittaels, K.; Van Den Haute, C.; Mercken, M.;Moechars, D.; Laenen, I.; Kuiperi, C.; et al. Prominent cerebral amyloid angiopathy in transgenic miceoverexpressing the london mutant of human APP in neurons. Am. J. Pathol. 2000, 157, 1283–1298. [CrossRef]

36. Wirths, O.; Multhaup, G.; Czech, C.; Feldmann, N.; Blanchard, V.; Tremp, G.; Beyreuther, K.; Pradier, L.;Bayer, T.A. Intraneuronal APP/A beta trafficking and plaque formation in beta-amyloid precursor proteinand presenilin-1 transgenic mice. Brain Pathol. 2002, 12, 275–286. [CrossRef] [PubMed]

37. Wiltfang, J.; Esselmann, H.; Cupers, P.; Neumann, M.; Kretzschmar, H.; Beyermann, M.; Schleuder, D.;Jahn, H.; Ruther, E.; Kornhuber, J.; et al. Elevation of beta-amyloid peptide 2-42 in sporadic and familialAlzheimer’s disease and its generation in PS1 knockout cells. J. Biol. Chem. 2001, 276, 42645–42657. [CrossRef]

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open accessarticle distributed under the terms and conditions of the Creative Commons Attribution(CC BY) license (http://creativecommons.org/licenses/by/4.0/).